# CorpusIQ
> AI intelligence layer for business data. Accelerate your business. All your tools in one place. Gives AI assistants and AI agents governed access to live business data through a single connection. Read-only external-source retrieval tools do not write back to vendor systems; CorpusIQ control-plane tools, the only tools that accept writes, are separately named and annotated and act on your CorpusIQ configuration. Ask a question in plain English and get source-cited answers from 40+ connected business systems at once. 150+ pre-built expert skills auto-fire on plain-English questions. Direct MCP does not retain raw customer files or full connector response payloads; scoped operational logs may persist for up to 30 days.
## Core Pages
- [Pricing](https://www.corpusiq.io/pricing): Solo $29.95/mo, Team $109.95/mo, Business $189.95/mo. All tiers include 40+ connectors and a 30-day free trial.
- [Pricing summary, Markdown](https://www.corpusiq.io/pricing.md): Plain-text plan, price, and user-limit table generated from the same canonical source as the pricing page.
- [Connectors](https://www.corpusiq.io/connectors): The full list of 40+ live business tools and the source-specific retrieval scopes each connector requests.
- [Features](https://www.corpusiq.io/features): How CorpusIQ delivers source-cited answers across the connected stack.
- [Live Demo](https://demo.corpusiq.io): Hosted demo environment to try CorpusIQ.
- [Skills](https://www.corpusiq.io/skills): The 150+ pre-built workflows, what each one answers, and which connectors it pulls from.
- [Security](https://www.corpusiq.io/security): Read-only external-source retrieval, separately annotated CorpusIQ control-plane tools, scoped retention by data class, and connector-disconnect controls.
- [Enterprise](https://www.corpusiq.io/enterprise): Single sign-on, audit logs, custom deployment, and procurement notes.
- [Affiliate Program](https://www.corpusiq.io/affiliate): Partners earn 25% commission every year for 3 years on referred subscriptions, plus $1.50 per qualified free trial signup.
- [Blog](https://www.corpusiq.io/blog): Connector guides, playbooks, and technical deep dives.
## What is CorpusIQ?
CorpusIQ is an AI intelligence layer for business data. It gives AI assistants and AI agents governed access to live business data through a single connection. Read-only external-source retrieval tools do not write back to vendor systems; CorpusIQ control-plane tools, the only tools that accept writes, are separately named and annotated and act on your CorpusIQ configuration. A user asks a question in plain English inside ChatGPT, Claude, or Perplexity. CorpusIQ reads multiple connected business systems simultaneously, validates the numbers against consistent KPI definitions, and returns a source-cited answer. No dashboards to build. No CSV exports. Source access remains user-authorized.
## How CorpusIQ is different from a connector platform
A connector platform moves data between systems. CorpusIQ is an intelligence layer that reads live data from multiple systems at once and reasons across all of it to answer a single question. The intelligence layer includes query understanding, a metrics registry, a source-of-truth engine, a validation engine, and audit citations. CorpusIQ does not build data pipelines. It answers questions.
## Supported AI Assistants
- Claude (Anthropic)
- ChatGPT (OpenAI)
- Perplexity
CorpusIQ also supports any API-based or self-hosted model through a direct MCP connection, and works inside Slack, with Microsoft Teams coming soon.
## ChatGPT App Store
CorpusIQ is an official app in the ChatGPT app store. Find it at https://chatgpt.com/apps?q=corpusiq
## Azure Marketplace
CorpusIQ is available as a SaaS listing on Microsoft Azure Marketplace. Find it at https://marketplace.microsoft.com/en-us/product/saas/corpusiq2011.corpusiq?tab=Overview
## Connectors (40+)
Gmail, Outlook, Google Calendar, Calendly, Slack, Google Drive, Google Docs, Google Sheets, OneDrive, Dropbox, Google Analytics 4, Google Ads, Meta Ads, PostHog, Google Search Console, Semrush, Ahrefs, HubSpot, HighLevel, Odoo, Monday.com, Klaviyo, Mailchimp, Constant Contact, ActiveCampaign, YouTube, TikTok, Facebook, Instagram, Shopify, Etsy, eBay, GunBroker, QuickBooks, Stripe, Airtable, PostgreSQL, SQL Server, MySQL, AWS S3, MongoDB, Azure Cosmos DB, and more.
## Key Features
- AI intelligence layer for business data, not a connector platform
- Cross-source reasoning: one question reads multiple connected systems at once
- Source-cited answers with links to the originating record
- Read-only external-source retrieval tools that do not write back to connected systems; CorpusIQ control-plane tools, the only tools that accept writes, are separately named and annotated and act on your CorpusIQ configuration
- Scoped retention by data class, with raw customer files excluded from direct-MCP retention
- 150+ pre-built expert skills, 650+ tools surfaced through skills
- Metrics validation against consistent KPI definitions
- Direct MCP excludes raw customer files and full connector response payloads from retention; CorpusIQ does not train models on customer data, while each AI provider's selected plan governs conversation handling
## Getting Started
1. Sign up at https://www.corpusiq.io
2. Connect your business tools via OAuth
3. Ask questions in Claude, ChatGPT, or Perplexity
## Top Skills
CorpusIQ's pre-built expert workflows are documented in dedicated spoke pages. Ten featured skill workflows:
- [Ad Spend Truth Report](https://www.corpusiq.io/skill/ad-spend-truth-report): Reconcile ad platform claims against GA4 and Shopify booked revenue. Catch over-attributed spend before doubling a budget.
- [Cash Recovery Engine](https://www.corpusiq.io/skill/cash-recovery-engine): Identify aged AR, draft collection emails per aging bucket, and estimate recoverable dollars this month from QuickBooks.
- [Financial Command Center](https://www.corpusiq.io/skill/financial-command-center): Single-screen weekly financial view from QuickBooks, Shopify, and the ad platforms.
- [Executive Snapshot](https://www.corpusiq.io/skill/executive-snapshot): Daily one-screen operating view with cash, revenue, ad efficiency, and the one signal worth attention.
- [Ecommerce Command Center](https://www.corpusiq.io/skill/ecommerce-command-center): Weekly ecommerce operating view across Shopify, GA4, and the paid channels with the top three actions.
- [Churn Prevention](https://www.corpusiq.io/skill/churn-prevention): Surface at-risk customers from usage decay and email engagement, with intervention plays per segment.
- [CRM Pipeline Health](https://www.corpusiq.io/skill/crm-pipeline-health): Diagnose pipeline health from CRM activity, calendar coverage, and quote-to-close patterns.
- [Customer Health Scorecard](https://www.corpusiq.io/skill/customer-health-scorecard): Score active customers red, yellow, green with the action each red customer needs this week.
- [SEO Audit](https://www.corpusiq.io/skill/seo-audit): Quarterly SEO audit grounded in Search Console with a prioritized 30-day fix list.
- [AI SEO](https://www.corpusiq.io/skill/ai-seo): Audit why a brand fails to surface in ChatGPT, Claude, and Perplexity for category prompts and produce a closeout plan.
## Learn More
- [What is MCP?](https://www.corpusiq.io/mcp)
- [Security Details](https://www.corpusiq.io/security)
- [Enterprise](https://www.corpusiq.io/enterprise)
## Full Version
- [llms-full.txt](https://www.corpusiq.io/llms-full.txt): Same header as this file, plus the full text of every published blog post for retrieval-grade ingestion.
---
## Latest Blog Posts (Full Content)
---
# How MCP Works: The Protocol Behind AI Tool Use
URL: https://www.corpusiq.io/blog/how-mcp-works-protocol-claude-chatgpt
Published: 2026-08-28
Category: tech-deep-dive
Connector: all
Model Context Protocol lets Claude, ChatGPT, and Perplexity call external tools. Here is how MCP works and why it matters for AI-native businesses.
import { ConnectorCallout } from "@/components/blog/ConnectorCallout";
import { FAQ } from "@/components/blog/FAQ";
The Model Context Protocol is the reason Claude can query your QuickBooks, ChatGPT can search your Gmail, and Perplexity can pull from your Shopify. Before MCP, every AI-tool integration was a one-off. After MCP, tools and models interoperate.
This post explains what MCP actually is, how it works at the protocol level, and why the architecture decisions matter for anyone evaluating AI-native tooling.
## The problem MCP solves
Before MCP, connecting an AI assistant to an external tool required a custom integration per model. An integration with Claude would not work with ChatGPT. A Gmail connector built for one model needed to be rebuilt for every other model.
The analogue is the web before standard protocols. Every browser implemented its own way of talking to servers. Once HTTP and HTML standardized, the web became interoperable.
MCP plays the same role for AI tool use. A server that exposes QuickBooks via MCP works with any MCP-compliant client: Claude, ChatGPT, Perplexity, or anything else. Build once, work everywhere.
## The three components of MCP
MCP defines three roles.
**Host.** The application the user talks to. Claude Desktop, Claude on the web, ChatGPT, Perplexity. The host presents the conversation UI and runs the model.
**Client.** The component inside the host that speaks MCP. Every host has an MCP client that handles the protocol mechanics: connecting to servers, discovering tools, routing calls.
**Server.** The component that exposes tools. This is where CorpusIQ lives. A CorpusIQ MCP server exposes QuickBooks, Shopify, Gmail, and 20+ other tools. The server describes each tool, its inputs, and its outputs. When the model wants to call a tool, the client routes the call to the server, the server executes it, and the result comes back.
## How a conversation actually flows
A user asks Claude, "what is my current cash position in QuickBooks?"
1. Claude's host sends the user message to the model.
2. The model recognizes this requires external data. It responds with a tool call: "use the QuickBooks tool `get_balance_sheet` with these parameters."
3. The host's MCP client receives the tool call intent and routes it to the CorpusIQ MCP server.
4. CorpusIQ's server receives the call, authenticates with QuickBooks using the user's OAuth token, executes the query against QuickBooks Online's API.
5. The QuickBooks response flows back through the server to the client to the model.
6. The model uses the data to compose a human-readable answer.
7. Claude shows the answer to the user.
Steps 1, 2, 6, and 7 happen inside Anthropic. Steps 3-5 happen between Anthropic's client and the CorpusIQ server. The user experiences it as a natural conversation.
## Discovery, capability negotiation, and tool definitions
The part that makes MCP extensible is discovery.
When a client connects to a server, the server sends a manifest. The manifest lists tools, their names, descriptions, input schemas, and output schemas. The model does not know what tools exist until this discovery happens.
This matters because the set of available tools changes based on what the user has connected. A user with QuickBooks gets QuickBooks tools. A user with Shopify gets Shopify tools. A user with both gets both sets, and Claude can call either.
The manifest is also how tool descriptions inform model behavior. A well-described tool ("Search Gmail for messages matching a query. Returns message IDs, sender, subject, and snippet.") gets called at the right times. A poorly-described tool gets called incorrectly or ignored.
## Why the transport layer matters
MCP uses JSON-RPC 2.0 over stdio or HTTP+SSE. That sounds boring; it is not.
JSON-RPC is a stateful protocol. Client and server maintain a connection. Either side can send messages at any time. This allows the model to call tools multiple times in a single conversation turn without reconnecting, and it allows servers to push notifications (for example, "your connection expired").
The stdio transport is for local MCP servers running on the same machine as the client. Useful for desktop apps.
The HTTP+SSE (Server-Sent Events) transport is for remote servers. CorpusIQ runs HTTP+SSE so Claude and ChatGPT in the cloud can reach CorpusIQ's servers wherever they are deployed.
## What CorpusIQ adds on top of MCP
Raw MCP is a protocol. It does not solve authentication, data safety, rate limiting, or scale.
CorpusIQ implements MCP and adds:
OAuth flows per connector. QuickBooks wants OAuth one way, Shopify another, Google a third. CorpusIQ handles each vendor's authentication idiosyncrasies and exposes a uniform MCP interface to the model.
Read-only enforcement. CorpusIQ's tool definitions expose only read operations. The model cannot trigger write actions through CorpusIQ regardless of what it tries.
Rate limiting. If a model tries to make 500 QuickBooks calls in a minute, CorpusIQ's auth gate enforces reasonable limits before any of those hit Intuit's servers.
Multi-connector queries. A user asks about a customer across QuickBooks, Gmail, and HubSpot. CorpusIQ's server exposes all three and Claude can call any of them in sequence to build the answer.
Managed infrastructure. CorpusIQ runs on Azure Container Apps with health checks, audit logging, scoped retention, and per-tenant token storage in Azure Key Vault. Direct MCP does not retain raw customer files or full connector response payloads; operational logs may persist for up to 30 days.
None of this is mandated by MCP. MCP is the protocol. CorpusIQ is the production-grade implementation.
## What MCP does not do
Three things.
MCP does not handle authentication. Each server implements its own auth. CorpusIQ handles OAuth for 40+ connectors; that is work MCP did not define.
MCP does not define data residency or retention policy. CorpusIQ applies separate lifecycles for direct retrieval, 30-day Azure Log Analytics logs, and optional indexed search. The selected AI client's plan and settings govern conversation handling after receipt.
MCP does not guarantee safety. A badly designed MCP server could expose dangerous tools to any connected model. Safety requires intentional design, which is why CorpusIQ enforces read-only scopes across all connectors.
## Why this matters for AI-native businesses
Two reasons to care.
First, MCP is the standard. Betting on a single-vendor tool integration is a strategy that dies the moment that vendor changes its API or you want to switch models. MCP-based integrations work with any MCP client, today or in the future.
Second, the reasoning layer can now cross tools freely. The old model (integrations pre-defined by the vendor, one workflow per tool) is replaced by: connect many tools, let the model reason across them. The compounding value is high.
## See also
- [Read-Only Retrieval and Scoped Data Handling: The CorpusIQ Security Architecture](https://www.corpusiq.io/blog/read-only-oauth-zero-data-storage-architecture)
- [Why We Built Mega-Tools: Collapsing 225 Tools to 24 for ChatGPT](https://www.corpusiq.io/blog/mega-tools-collapsing-225-to-24-chatgpt)
- [Portable Data: Why CorpusIQ Ships a .cnw.zip Export Format](https://www.corpusiq.io/blog/portable-data-cnw-zip-export-format)
---
# Vendor Tracker: Spot Overpayments Before the Audit
URL: https://www.corpusiq.io/blog/vendor-relationship-tracker-overpayments
Published: 2026-08-24
Category: use-case
Connector: quickbooks
Vendor spend drifts and nobody notices until the audit. Connect QuickBooks, Drive, and email to Claude with CorpusIQ for a monthly vendor health review.
import { ConnectorCallout } from '@/components/blog/ConnectorCallout'
import { FAQ } from '@/components/blog/FAQ'
Vendor relationships decay invisibly. The agency you hired two years ago still gets paid $8,000 a month; nobody has checked what they deliver in six months. The SaaS tool that was central in 2023 is used by two people now but still renews every year. The supplier you pay on net-30 now pays you on net-60 and nobody renegotiated.
Connecting QuickBooks, Google Drive, and email to Claude through CorpusIQ makes vendor health a 30-minute monthly review instead of an annual audit that never quite happens.
## The three patterns that cost money
Most vendor waste falls into three buckets.
Duplicate or overlapping services. Two project management tools. Two analytics platforms. Three email marketing tools across different departments. Nobody decided to have them all; they accumulated.
Unused subscriptions. A tool that was evaluated and left running. The team moved on. The subscription renews. Sometimes for years.
Drifted pricing. Vendor raised prices quietly. Operator did not notice because the change was gradual or buried in a renewal.
The total is usually 3-8% of vendor spend. At $10M in revenue with $2M in vendor spend, that is $60k-$160k a year recoverable.
## The monthly vendor review
Run one prompt a month.
```
Pull the top 50 vendors by total paid amount in QuickBooks over the last 12 months.
For each vendor:
1. Monthly or annual spend pattern
2. Most recent contract or invoice in Drive
3. Last 90 days of email with this vendor (response cadence, any tone change,
any issues raised)
4. Any payments that look anomalous: doubled months, missed months, sudden
increases
Produce:
- Vendors where spend has increased more than 20% without obvious justification
- Vendors where we have not had email contact in the last 60 days (dying
relationship or pure subscription renewal)
- Vendors with contracts that expire in the next 120 days
- Any duplicate services (two vendors in the same category)
- Any SaaS subscription with low apparent usage based on internal mentions in
email
```
Claude returns the health list. You review, investigate, and act.
## What action looks like
Three actions per flagged vendor.
Keep: service is essential, price is fair, relationship is healthy. No action.
Renegotiate: service is essential but price has drifted. Use the contract expiration window or the next payment cycle to ask for a discount or an adjusted scope. Data from QuickBooks shows exactly what you have paid; contract from Drive shows what was promised.
Cancel: service is not essential, or is duplicative with another vendor. Cancel and notify before the next renewal window.
Three actions is all. The skill is being decisive. The default of "leave it alone" is what caused the drift.
## The case for email as a signal
Payment data alone misses the softest signal: the relationship is dying.
A healthy vendor relationship has regular email. QBRs, check-ins, occasional problem-solving, occasional price discussions. When the email goes quiet, one of three things is true: the vendor has stopped caring, your team has stopped caring, or the tool is pure infrastructure you do not need to discuss.
Any of those is a signal. Claude can surface which is which by reading tone and cadence across the last 90 days.
## How to set it up
1. Sign up at [corpusiq.io](https://www.corpusiq.io). Solo $29.95/month.
2. Connect QuickBooks Online.
3. Connect Google Drive or Dropbox for contracts.
4. Connect Gmail or Outlook.
5. Add CorpusIQ MCP to Claude.
## Sample prompts for specific vendor questions
- "For vendor [NAME], show me the 12-month payment history and flag any anomalies."
- "Which vendors have had more than 2 invoices rejected or disputed in the last year?"
- "Find vendors we pay monthly but have not sent us an invoice in the last 90 days."
- "Compare the contract price for [VENDOR] in Drive against what we actually paid last year."
- "List every SaaS subscription we pay over $500/month. For each, summarize what the tool does and whether it appears in Slack or email discussions."
- "Which vendors have annual auto-renewals coming up in the next 90 days?"
- "Find any vendor where the last email exchange mentions a price increase."
## What this does not catch
Some waste requires process, not software.
A service that is genuinely under-used but not surfaced in email or contract scans stays invisible. Someone has to actually compare "what we pay" to "what we use" for some subscriptions, and Claude cannot measure usage for every tool.
Off-contract procurement. If a team is paying for a service via a personal credit card rather than through QuickBooks AP, Claude cannot see it. Audit your expense reimbursement process separately.
The vendor review process catches most of the structured waste. It is a high-ROI activity that CorpusIQ makes fast enough to actually run.
## See also
- [Contract Intelligence Suite: Never Miss a Renewal Again](https://www.corpusiq.io/blog/contract-intelligence-never-miss-renewal)
- [Connect QuickBooks to Claude: Close Your Books 4x Faster](https://www.corpusiq.io/blog/connect-quickbooks-claude-close-books-faster)
- [Scope Drift Detection: Catch Revenue Leakage Across Drive, Email, and QuickBooks](https://www.corpusiq.io/blog/scope-drift-detection-revenue-leakage)
---
# Ad Spend Truth: True ROAS Across Google, Meta, TikTok
URL: https://www.corpusiq.io/blog/ad-spend-truth-report-true-roas
Published: 2026-08-20
Category: use-case
Connector: google-ads
Every ad platform reports its own ROAS. Connect Google Ads, Meta, TikTok, GA4, and Shopify to Claude with CorpusIQ for blended ROAS and real CAC.
import { ConnectorCallout } from '@/components/blog/ConnectorCallout'
import { FAQ } from '@/components/blog/FAQ'
Every ad platform reports its own ROAS. Google Ads says 4.2. Meta says 3.8. TikTok says 5.1. Summed, they claim to have generated more revenue than your store actually produced. That is not an accounting error. It is the platforms each taking credit for the same conversions.
Operators who believe platform-reported ROAS overspend. Operators who know the gap, adjust. Connecting Google Ads, Meta, TikTok, GA4, and Shopify to Claude through CorpusIQ gives you the blended view in minutes instead of building a Northbeam subscription around it.
## Why platform ROAS is misleading
Three mechanisms inflate it.
View-through conversion. Meta counts a sale if the user saw an ad in the last 1-7 days, even if they did not click. If that user also clicked a Google ad, Google also counts it.
Attribution window generosity. Most platforms use a 7-day click and 1-day view window by default. Longer windows capture more conversions per dollar, making ROAS look better.
Self-reporting incentive. Every platform is incentivized to show high ROAS. They set defaults that are favorable to themselves.
The consequence: operators read three platform reports, think ads are working, spend more. The actual revenue does not grow as fast as the reported ROAS suggests it should.
## The truth report
Two metrics cut through the attribution noise.
MER (Marketing Efficiency Ratio): total revenue divided by total ad spend, across all platforms. Does not rely on attribution at all. Treats the whole ad budget as one investment and measures whether revenue follows.
Incremental blended ROAS: total ad spend divided into total revenue from all paid sources per GA4. More precise than MER but depends on GA4 attribution, which has its own issues.
For most brands, MER is the cleanest weekly number. Track it over time. Trend matters more than absolute value.
## The weekly truth report prompt
```
Pull the following for last week:
From Shopify: total revenue, order count, AOV, new customer count.
From Google Ads: total spend, platform-reported conversions and revenue.
From Meta Ads: total spend, platform-reported conversions and revenue.
From TikTok Ads: total spend, platform-reported conversions and revenue (if connected).
From GA4: paid traffic sessions, paid conversions, paid conversion revenue.
Calculate:
1. Total ad spend across all platforms
2. MER (Shopify revenue / total ad spend)
3. Platform-reported ROAS per channel (platform revenue / platform spend)
4. Sum of platform-reported revenue versus Shopify total revenue
5. GA4 paid conversion revenue versus sum of platform-reported
Highlight:
- The gap between summed platform ROAS and MER
- Any platform where platform-reported ROAS has diverged from GA4-attributed
ROAS over the last 4 weeks
- MER trend compared to the prior 4 weeks
```
Claude synthesizes across five connectors. Output is a clear truth report in five minutes instead of an afternoon of spreadsheet work.
## What to do with the findings
Four actions flow from the truth report.
First, set an MER target. Most ecommerce brands target 3.0-5.0 depending on margin structure. Watch weekly. If MER compresses below target, the ad budget is not earning its keep.
Second, identify the platform with the biggest attribution inflation. If Google Ads claims $50k and GA4 shows $30k from paid search, Google's self-reporting is suspect. That does not mean stop spending on Google. It means discount the reported ROAS.
Third, run holdout tests. The only true way to know incremental ad impact is to pause a channel for a week and measure. Claude can quantify the revenue impact of the holdout by comparing to the prior period.
Fourth, reallocate. Channels with the most inflated attribution are often the ones to cut first when MER compresses. The non-incremental portion of their spend was waste.
## How to set it up
1. Sign up at [corpusiq.io](https://www.corpusiq.io). Solo $29.95/month covers all the connectors needed.
2. Connect Shopify, Google Ads, Facebook Marketing, TikTok (if applicable), and Google Workspace for GA4.
3. Add CorpusIQ MCP to Claude.
## What this does not replace
Two things.
Advanced attribution modeling. Tools like Northbeam, Triple Whale, and Wicked Reports build media mix models using statistical techniques. This workflow is directional, not rigorous. For brands over $50M ad spend, proper MMM is worth the investment.
Creative-level optimization. The truth report tells you whether the whole ad budget is working. It does not tell you which creative is working. Use the platform ad-level reports for that, with the caveat that attribution there is also self-reported.
The value of the truth report is not perfect accuracy. It is seeing the scale of the attribution inflation and treating platform ROAS as a signal, not gospel.
## See also
- [Connect Google Ads to Claude: Find Wasted Spend in Minutes](https://www.corpusiq.io/blog/connect-google-ads-claude-find-wasted-spend)
- [Connect Facebook Ads to ChatGPT: Campaign Analysis Without the Ads Manager](https://www.corpusiq.io/blog/connect-facebook-ads-chatgpt-campaign-analysis)
- [Weekly Business Pulse Without a BI Team: GA4 + Shopify + Ads in One Report](https://www.corpusiq.io/blog/weekly-business-pulse-ga4-shopify-ads)
---
# Contract Intelligence Suite: Never Miss a Renewal Again
URL: https://www.corpusiq.io/blog/contract-intelligence-never-miss-renewal
Published: 2026-08-13
Category: use-case
Connector: google-drive
Missed renewals cost real money. Connect Drive, Dropbox, and email to Claude with CorpusIQ for a renewal radar on every auto-renewal and notice window.
import { ConnectorCallout } from '@/components/blog/ConnectorCallout'
import { FAQ } from '@/components/blog/FAQ'
Every company signs contracts and then forgets what they say. The SaaS subscription that auto-renews in June with a 60-day notice window. The vendor agreement that escalates 5% annually without anyone reading the clause. The MSA whose exclusivity terms would have been useful to know before signing a competing deal.
The pattern is universal. Sign the contract, save it, revisit only when something breaks. By then, the renewal has happened, the price has escalated, or the term has committed you.
Connecting Drive, Dropbox, and email to Claude through CorpusIQ turns contract monitoring from a "someone should do this" into a monthly 15-minute audit.
## Why contract drift goes unmanaged
Three reasons.
Contracts are not one system. They are signed in DocuSign or PandaDoc, filed in Drive or Dropbox, referenced in emails, and sometimes summarized in Notion. No single location has everything.
Reading contracts is work. Most are 15-40 pages. Nobody reads them quarterly for fun. The sharp clauses (auto-renewal, escalation, termination notice) are usually deep in the document.
Calendar reminders decay. Someone sets a reminder when they sign. The reminder fires 18 months later. The person who set it has left the company, or has forgotten why the reminder matters.
## The monthly contract audit
Connect the systems where contracts live. Run one prompt a month.
```
Search Drive, Dropbox, and email attachments for all contracts and agreements.
For each:
- Counterparty
- Signed date and term
- Expiration date
- Auto-renewal clause: does it auto-renew, and what is the notice window?
- Termination rights: what is required to exit?
- Price escalation clause
- Any unusual obligations (exclusivity, minimum spend, performance
guarantees)
Then produce:
1. Contracts expiring in the next 180 days
2. Contracts with auto-renewal where the notice window opens in the next 90 days
3. Contracts with annual price escalations due in the next 90 days
4. Contracts with minimum spend commitments we may not hit
```
Claude returns the list. You act on what needs action.
## What "action" looks like
For each flagged contract:
Contracts expiring soon: decide renew, renegotiate, or let lapse. The earlier you decide, the better your leverage.
Auto-renewals with notice window opening: decide if you want to renew. If not, send the termination notice inside the window. Notice windows are often the only leverage moment. Missing them locks you in.
Price escalations: review whether the escalated price is still competitive. Compare to current market. Use as a renegotiation anchor.
Minimum spend commitments: check current-year spend against commitment. If tracking to miss, plan either to absorb the penalty or ramp spend intentionally.
## How to set it up
1. Sign up at [corpusiq.io](https://www.corpusiq.io). Solo $29.95/month, includes Drive, Dropbox, Gmail, and 19 more connectors.
2. Connect every system where contracts live. For most SMBs: Google Drive and Gmail or Outlook.
3. Add CorpusIQ MCP to Claude.
## Sample prompts for specific contract questions
- "Find all SaaS contracts in Drive. For each, list monthly cost, contract term, and renewal date."
- "Search for every agreement with a termination-for-convenience clause. What notice period does each require?"
- "Which vendor contracts have escalation clauses over 5% per year?"
- "Find any agreement that references a minimum spend commitment. What is each commitment and have we hit it this year?"
- "Search contracts for non-compete or exclusivity language. List what each restricts."
- "Which contracts signed in the last 12 months have confidentiality clauses surviving past contract end?"
- "Find every agreement with a most-favored-nation clause."
- "Search contracts for liquidated damages or penalty clauses."
## The renewal negotiation playbook
Catching the renewal is step one. Using it is step two.
When a renewal window opens, ask Claude: "For this contract, what was the original pricing, what was the escalation history, and what does the current price compare to the market?" (The market comparison still requires external research, but Claude can structure what to look for.)
Go into the renewal conversation with data, not feelings. Vendor should know that you read their contract, tracked escalations, and are prepared to leave. That posture compresses their margin without damaging the relationship.
## What this does not do
Two caveats.
Claude cannot read scanned contracts without OCR. If your contract archive is full of scans, you need to OCR them first. Modern Drive and Dropbox often OCR automatically, but verify.
Claude cannot negotiate on your behalf. It surfaces information. The human conversations still require a human.
The workflow turns contract awareness from reactive to proactive. That shift alone saves most businesses more than the cost of CorpusIQ for years.
## See also
- [Connect Dropbox to Claude: Contract Intelligence Across Every Folder](https://www.corpusiq.io/blog/connect-dropbox-claude-contract-intelligence)
- [Vendor Relationship Tracker: Spot Overpayments and Dying Relationships](https://www.corpusiq.io/blog/vendor-relationship-tracker-overpayments)
- [Connect Google Drive to Claude: Turn Every Doc into Answerable Data](https://www.corpusiq.io/blog/connect-google-drive-claude-answerable-data)
---
# SaaS Pipeline Velocity with HubSpot, Gmail, and Calendar
URL: https://www.corpusiq.io/blog/saas-cro-pipeline-velocity-hubspot-gmail
Published: 2026-08-10
Category: use-case
Connector: hubspot
Pipeline velocity predicts the SaaS quarter. Connect HubSpot, Gmail, and Calendar to Claude with CorpusIQ for forecasts, stall detection, and rep performance.
import { ConnectorCallout } from '@/components/blog/ConnectorCallout'
import { FAQ } from '@/components/blog/FAQ'
Pipeline velocity is the most predictive number in SaaS. Revenue follows velocity with a lag of one sales cycle. If velocity is compressing, the next quarter is at risk, and knowing that early is the difference between adjusting and missing.
Most CROs watch pipeline stage count and top-of-funnel volume. They should also watch velocity: the rate at which deals move from stage to stage, the average cycle length by segment, the gap between email engagement and meeting booking. Connecting HubSpot, Gmail, and Google Calendar to Claude through CorpusIQ makes the full velocity view accessible without a RevOps build-out.
## What CROs actually need to know weekly
Five questions. Most SaaS leadership teams can answer one or two cleanly. The rest live in spreadsheets nobody wants to maintain.
Which deals in the forecast are at risk, and why? Stage plus age plus activity. A deal that has been in "proposal sent" for 21 days with no email reply is different from one that was sent yesterday.
Which reps are compressing cycle time this quarter versus last, and which are extending? Average days-in-stage by rep, trended.
What is the real conversion rate from meeting booked to opportunity created, by source? Not what HubSpot says, what calendar plus CRM says.
Which enterprise deals have not had a meeting scheduled in the last 21 days? These are deals sliding quietly.
Which top-of-funnel activities (emails sent, meetings booked) are trending the wrong direction this month?
## The velocity dashboard without a dashboard
Connect HubSpot, Gmail, and Calendar. Run a single prompt weekly.
```
Give me a pipeline velocity review for this week.
From HubSpot:
- Deal count by stage
- Average days in current stage by stage
- Close date slippage: deals where close date has moved more than once
- Weighted pipeline for the next 90 days
- Average cycle time for closed-won deals in the last 90 days versus the
prior 90 days
From Gmail:
- Number of emails sent to prospects this week (by rep if possible)
- Open enterprise deals with no outbound email in the last 14 days
From Google Calendar:
- External meetings booked this week, grouped by rep
- Comparison to last week
- Any open deal with no scheduled meeting in the next 21 days
Produce a one-page CRO review with:
1. Headline: is velocity up or down week over week
2. Three specific deals at risk and why
3. Two trends worth watching
```
Claude returns the synthesis. You read for five minutes, forward to the sales leader, discuss at the pipeline meeting.
## Deal-at-risk detection
The hardest thing a SaaS CRO does is tell which deals in the forecast will close and which will not. Gut feel works for small pipelines. It fails as the portfolio grows.
For each enterprise deal in the forecast, ask Claude to run:
```
For deal [DEAL NAME] in HubSpot, pull:
- Current stage and days in stage
- Close date and any prior changes to close date
- Last email exchange with the primary contact
- Last meeting on calendar and next scheduled meeting (if any)
- Any activity in Slack referencing this account in the last 30 days
Assess the risk. Is this deal tracking to close, slipping, or stalled?
```
Claude synthesizes across three connectors and gives an informed view. The RevOps analysis that used to take 20 minutes per deal takes two.
## The weekly cadence
Monday morning: run the velocity review prompt. Read the summary. Identify the three highest-risk deals.
Monday pipeline meeting: the sales leader has Claude's output as the baseline. Each rep presents their deals with Claude's risk assessment as context. The meeting is about what to do, not about whose forecast is right.
Mid-week: run deal-at-risk prompts on any deal where the rep disagrees with Claude's assessment. Dig in.
Friday: quick follow-up prompt on any deal flagged earlier in the week. Did activity pick up? Is the deal actually moving?
This rhythm is how a CRO converts noisy CRM data into real forecast accuracy.
## How to set it up
1. Sign up at [corpusiq.io](https://www.corpusiq.io). Solo $29.95/month, Team at $109.95/month if 5 users need access.
2. Connect HubSpot (or another CRM as available).
3. Connect Google Workspace for Gmail and Calendar.
4. Add CorpusIQ MCP to Claude.
## What this does not fix
Two honest caveats.
It does not fix bad CRM hygiene. If your reps do not update deal stages and do not log activities, Claude has nothing to reason over. Hygiene is upstream and stays upstream.
It does not replace sales skill. Claude flags deals at risk. Closing deals at risk still takes a good salesperson doing the right work. The tool makes the diagnosis faster; the treatment is still human.
## See also
- [Connect HubSpot to Claude: Pipeline Health in One Conversation](https://www.corpusiq.io/blog/connect-hubspot-claude-pipeline-health)
- [CRM Pipeline Health Monitor](https://www.corpusiq.io/blog/connect-hubspot-claude-pipeline-health)
- [Connect Gmail to Claude: Recover Critical Emails Before They Cost You](https://www.corpusiq.io/blog/connect-gmail-claude-recover-critical-emails)
---
# Real Estate Portfolio Intelligence with ChatGPT
URL: https://www.corpusiq.io/blog/real-estate-portfolio-intelligence-chatgpt
Published: 2026-08-03
Category: use-case
Connector: google-drive
Real estate runs on paperwork. Connect Drive, QuickBooks, and email to ChatGPT with CorpusIQ for lease reviews, rent roll analysis, and vendor oversight.
import { ConnectorCallout } from '@/components/blog/ConnectorCallout'
import { FAQ } from '@/components/blog/FAQ'
Real estate operators juggle paperwork at a scale other industries do not. Every property has a lease (sometimes multiple). Every vendor has a service contract. Every unit has a rent roll, a maintenance history, and a payment history. Finding specific information means hunting through a property management system, files in Drive or Dropbox, emails with tenants and vendors, and payment records in QuickBooks.
Connecting Google Drive, QuickBooks, and Gmail or Outlook to ChatGPT through CorpusIQ turns the paperwork into a queryable source. Ask about lease terms, rent roll anomalies, or vendor obligations. Answer pulls from the actual documents.
## Where real estate ops get stuck
Three recurring issues.
Lease discovery. Every property has a lease or several amendments. Finding the specific clause that governs a tenant dispute, a rent escalation, or a termination right takes 20 minutes of reading.
Rent roll anomalies. The rent roll looks fine at the summary level. Specific anomalies (a tenant paying below their contracted rate, a unit not billing late fees correctly, a lease that escalated but invoicing did not catch up) are invisible without drilling in.
Vendor oversight. Landscaping, HVAC, security, pest control. Every property has five to twenty vendor contracts. Missed renewal windows, overlapping services, and pricing drift go unnoticed until year-end.
## The Claude workflow for real estate portfolios
Four workflows earn the setup time.
**Quarterly lease audit.** Pull every active lease from Drive. Ask ChatGPT to summarize key terms: term length, rent amount, escalation clause, renewal options, notice requirements. Flag anything unusual.
**Rent roll reconciliation.** Compare the rent roll (in Drive or property management export) against actual receipts in QuickBooks. Flag any tenant whose received amount differs from the contracted amount.
**Vendor contract review.** Pull every vendor contract from Drive. Cross-reference against payments in QuickBooks. Flag contracts that have expired, contracts with auto-renewals approaching, and payments without a corresponding contract.
**Tenant issue triage.** Periodically search Gmail for tenant complaints, maintenance requests, and legal notices. Cross-reference against the tenant's payment history. Identify escalation risk.
## The quarterly lease audit prompt
```
Pull every lease and lease amendment from the Leases folder in Drive.
For each lease, summarize:
- Tenant name
- Property/unit
- Lease term (start and end date)
- Current monthly rent
- Rent escalation clause
- Renewal option (if any)
- Termination rights and notice requirements
- Any unusual clauses worth flagging
Then list:
- Leases expiring in the next 180 days
- Leases with escalations due in the next 90 days
- Any leases where the terms look non-standard for our portfolio
```
ChatGPT produces the summary. The operator has a clear action list in minutes.
## How to set it up
1. Sign up at [corpusiq.io](https://www.corpusiq.io). Solo $29.95/month.
2. Connect Google Drive (or Dropbox, or OneDrive) for leases and vendor contracts.
3. Connect QuickBooks Online for rent receipts and vendor payments.
4. Connect Gmail or Outlook for tenant and vendor communication.
5. Add CorpusIQ to ChatGPT.
## Sample prompts for real portfolio work
- "List every tenant whose rent escalation is due in the next 90 days. What should the new rent be based on the escalation clause?"
- "Which units in the portfolio have had more than 3 maintenance requests in the last 6 months?"
- "Compare the rent roll to QuickBooks receipts for last month. Flag any unit where the amounts do not match."
- "Find every vendor contract set to auto-renew in the next 60 days and summarize the renewal terms."
- "Which tenants have emailed us about rent concerns in the last 90 days?"
- "List vendor payments over $5,000 in the last quarter that do not have a corresponding contract in Drive."
- "What is the average vacancy duration across the portfolio in the last 12 months?"
- "Find every lease with a co-tenancy clause and summarize the triggers."
## Rent roll reconciliation in practice
The reconciliation step is where real money surfaces. The pattern:
Run the prompt "compare the rent roll to QuickBooks receipts for last month, unit by unit."
ChatGPT lists units where expected rent differs from received. Common causes:
- Tenant on a payment plan that was not updated in the rent roll
- Missed late fee
- Partial payment that was not followed up
- Billing error (tenant charged old rate after escalation)
- Rent concession that was not documented in the tenant ledger
Most of these are recoverable. The time to recover them compresses from a month-long audit to an afternoon.
## What stays outside the system
Property management platforms (AppFolio, Yardi, Buildium) are not yet direct connectors for most CorpusIQ deployments. The workaround: export rent roll and tenant ledger to Drive regularly, and ChatGPT can read the exports.
Physical property inspection data rarely lives in queryable systems. Maintenance logs in a clipboard on a property manager's desk are invisible to ChatGPT.
## See also
- [Contract Intelligence Suite: Never Miss a Renewal Again](https://www.corpusiq.io/blog/contract-intelligence-never-miss-renewal)
- [Vendor Relationship Tracker: Spot Overpayments and Dying Relationships](https://www.corpusiq.io/blog/vendor-relationship-tracker-overpayments)
- [Connect QuickBooks to Claude: Close Your Books 4x Faster](https://www.corpusiq.io/blog/connect-quickbooks-claude-close-books-faster)
---
# Track Supplier Risk Across QuickBooks, Drive, and Email
URL: https://www.corpusiq.io/blog/manufacturing-ops-supplier-risk-tracking
Published: 2026-07-27
Category: use-case
Connector: quickbooks
Manufacturing runs on paper trails. Connect QuickBooks, Drive, and email to Claude with CorpusIQ for supplier risk, PO tracking, and audit prep.
import { ConnectorCallout } from '@/components/blog/ConnectorCallout'
import { FAQ } from '@/components/blog/FAQ'
Manufacturing is a documentation-heavy business. Purchase orders in one system, supplier contracts in Drive, quality correspondence in email, payments in QuickBooks. When something goes wrong (a late delivery, a quality issue, a compliance question), the evidence is scattered across systems that do not talk to each other.
Connecting QuickBooks, Google Drive, and Gmail or Outlook to Claude through CorpusIQ gives manufacturing operators a single lens across the paper trail. Ask Claude about supplier risk, PO status, or compliance evidence. Answers pull from all three sources in seconds.
## Where manufacturing ops lose time
Three recurring pains.
Supplier risk surveillance is reactive, not proactive. The business learns a supplier is in trouble when the shipment does not arrive. The warning signs were visible earlier: slower responses to emails, delayed invoices, longer lead times. But nobody was watching across sources.
Audit prep is brutal. ISO, customer quality audits, government contract reviews. Each one requires pulling evidence from multiple systems. A week of ops time per audit is normal.
PO status is opaque. Where is the $80,000 casting order placed with Supplier X? The answer involves QuickBooks (payment status), Drive (PO and contract), and email (shipment updates). Stitching together takes a phone call or two.
## The Claude-assisted manufacturing workflow
Three workflows earn the setup time.
**Weekly supplier health check.** Ask Claude to pull the top 20 suppliers by spend from QuickBooks, then for each one, summarize payment history (any slowdowns), Drive documents (any recent contract amendments or quality issues), and email threads (tone, responsiveness, any reported problems). Output is a ranked risk list.
**PO status on demand.** Ask Claude where any open PO stands. Claude queries QuickBooks for payment status, Drive for the PO and delivery terms, and email for shipment updates. Answer comes back in one paragraph.
**Audit evidence pull.** When a customer or regulator asks for evidence on a specific requirement, ask Claude to pull the relevant documents, emails, and payment records. Saves days of searching.
## The weekly supplier health check prompt
```
Pull the top 20 vendors by total spend in QuickBooks over the last 12 months.
For each vendor:
1. Summarize payment history. Flag any slowdowns in our payments to them or
theirs to us if reversed.
2. Search Drive for supplier agreements, quality documents, or recent amendments.
3. Search email for threads with this vendor in the last 90 days. Flag any
that mention delays, quality issues, pricing changes, or capacity concerns.
Produce a ranked risk list. Top of the list = most concerning based on the
combined signal.
```
Claude returns the list. You scan. The signals that would have taken a week to compile are visible in minutes.
## How to set it up
1. Sign up at [corpusiq.io](https://www.corpusiq.io). Solo $29.95/month covers all three connectors plus 20 more.
2. Connect QuickBooks Online, Google Workspace (for Drive + Gmail), or Microsoft 365 (for OneDrive + Outlook).
3. Add CorpusIQ MCP to Claude.
Add PostgreSQL or MSSQL if your production system writes to a database. Claude can then cross-reference production data against purchase orders and financials.
## Sample prompts for real manufacturing questions
- "What is the current status of PO 14782 placed with Acme Castings? Include payment, delivery schedule, and any email correspondence."
- "Which suppliers in the last 6 months have been consistently late on delivery based on email thread patterns?"
- "Pull all supplier contracts in Drive that expire in the next 120 days. Summarize termination and renewal terms."
- "Find the quality certification for Material Lot X. What lab tested it, when, and what were the results?"
- "Which suppliers have we paid late more than twice in the last 12 months? This affects our negotiation leverage."
- "What evidence do we have that we performed incoming inspection on shipments from Supplier Y in Q1?"
- "List every government contract obligation in the DFARS compliance folder, with the current status of each."
- "Which suppliers have sent us pricing increase notices in the last 90 days?"
## Audit prep as a conversation
The hardest thing about audits is not the work during the audit. It is the pre-audit discovery: what do we have, where is it, and does it match?
Instead of a week of evidence gathering, run Claude through the audit checklist item by item. For each requirement, ask where the evidence is. Claude pulls it from the relevant sources and flags gaps. The audit becomes a conversation about what is missing, not about what is there.
## What stays outside the system
Claude surfaces what exists in the connected sources. Some manufacturing-critical data lives elsewhere.
Shop floor systems (MES, SCADA) are not typical connectors. Data from these systems must be exported to Drive or a database Claude can access.
Physical inspection records that were never digitized cannot be queried. If your quality records are on paper, they need to be scanned (with OCR) before Claude can use them.
Supplier portal data is usually siloed. If your customer requires portal access for their supplier scorecards, you still have to log in and read it.
The point of this workflow is not to replace every manufacturing system. It is to stitch together the ones you already use for financial, documentation, and communication purposes.
## See also
- [Vendor Relationship Tracker: Spot Overpayments and Dying Relationships](https://www.corpusiq.io/blog/vendor-relationship-tracker-overpayments)
- [Connect QuickBooks to Claude: Close Your Books 4x Faster](https://www.corpusiq.io/blog/connect-quickbooks-claude-close-books-faster)
- [Regulatory Compliance Sentinel: Monitor Audit-Ready Operations](https://www.corpusiq.io/blog/audit-ready-ai-structured-logging)
---
# Cash Recovery in 14 Days: Turn AR into a Claude Conversation
URL: https://www.corpusiq.io/blog/cash-recovery-14-days-ar-claude
Published: 2026-07-23
Category: use-case
Connector: quickbooks
Most overdue AR is recoverable with focused follow-up. Connect QuickBooks, Gmail, and Drive to Claude with CorpusIQ for a 14-day cash recovery sprint.
import { ConnectorCallout } from '@/components/blog/ConnectorCallout'
import { FAQ } from '@/components/blog/FAQ'
Most small businesses have six-figure accounts receivable balances that could be cash in the bank inside two weeks. The blocker is not that clients will not pay. It is that nobody is running a disciplined follow-up process.
Cash recovery does not require a collections department. It requires a structured sprint: identify overdue invoices, understand each customer's context, send the right message at the right cadence, follow through. Connecting QuickBooks, Gmail, and Drive to Claude through CorpusIQ makes all four steps fast enough to run in an afternoon.
## Why AR stays stuck
Three reasons.
First, AR is invisible. The P&L shows revenue when invoiced. The business feels like it is earning. Nobody is watching the cash conversion cycle.
Second, collections feel uncomfortable. The person sending the email is often the same person who has to maintain the client relationship. The email gets delayed, softened, or skipped.
Third, follow-up is ad-hoc. Somebody nudges a few customers one week, then forgets for a month. The customers who get nudged pay. The ones who do not, do not.
Money that should be in the bank sits in QuickBooks as an open invoice. Multiply by 30 clients and the working capital impact is real.
## The 14-day recovery sprint
A structured cash recovery sprint looks like this.
**Day 0: inventory.** Pull every overdue invoice. Segment by age: 0-30, 30-60, 60-90, 90+ days past due.
**Day 1-2: context gathering.** For each overdue customer, look up payment history, contract terms, recent email threads, and any disputes. This is where Claude saves the most time.
**Day 3-5: first-touch follow-up.** Personalized email to every overdue account. Not a mass template. A short message that acknowledges the relationship and asks for payment or an update.
**Day 6-10: response handling.** Replies come in. Some pay immediately. Some ask for payment plans. Some raise disputes. Handle each.
**Day 11-14: escalation.** Unresponsive accounts get a second message, this one with a clear next step (payment plan, late fee, account pause).
By day 14, the majority of the under-90-day AR has either been paid, scheduled, or escalated. That is the win.
## The Claude workflow per customer
For each overdue customer, ask Claude:
```
Pull the open AR balance for [CUSTOMER NAME] from QuickBooks, including invoice
numbers, amounts, and original due dates.
Pull the last 12 months of payment history for this customer. Are they normally
on-time, late, or inconsistent?
Find the most recent contract or MSA for this customer in Drive. Summarize
the payment terms.
Search Gmail for the last 90 days of emails with this customer. Flag any
mention of payment, disputes, or service issues.
Based on all of the above, draft a follow-up email. Tone should match their
history: firm for chronic late payers, gentle for normally-prompt customers
who have one overdue invoice. Include specific invoice numbers and amounts.
```
Claude synthesizes across four sources and produces a draft. You read, adjust, send. Five minutes per customer instead of thirty.
## How to set it up
1. Sign up at [corpusiq.io](https://www.corpusiq.io). Solo $29.95/month covers QuickBooks, Gmail, and Drive.
2. Connect QuickBooks Online (for AR), Google Workspace (for Gmail and Drive).
3. Add CorpusIQ MCP to Claude.
## What Claude writes well, what needs your judgment
Claude writes good first drafts. It is weaker at three things.
Reading the relationship. If a customer is a strategic account where payment is stuck in their AP process (not a refusal to pay), the email needs a different tone than what Claude defaults to. Adjust.
Knowing when to escalate. If a customer has a 120-day-past-due balance and has ignored three emails, the next step is probably a call or a letter, not another email. Claude can draft the letter; you decide when the moment is.
Catching legal and contractual nuance. Some contracts have specific late payment clauses, dispute resolution steps, or arbitration requirements. Claude surfaces the contract text; you interpret it.
The workflow is force-multiplier, not replacement. The multiplier is significant: a solo ops lead can run a cash recovery sprint across 30 customers in a week, something that used to require three weeks.
## Measuring the impact
Track two numbers before and after the sprint.
Days sales outstanding (DSO). Total AR divided by average daily revenue. A 14-day sprint should drop DSO by 5-15%.
Percentage of AR over 60 days past due. This should compress meaningfully. If it does not, you have discovery work to do: disputes you did not know about, customers in genuine financial trouble, or invoicing errors on your side.
## See also
- [Connect QuickBooks to Claude: Close Your Books 4x Faster](https://www.corpusiq.io/blog/connect-quickbooks-claude-close-books-faster)
- [Close the Books 50% Faster: QuickBooks + Drive + Gmail Workflow with Claude](https://www.corpusiq.io/blog/close-books-faster-quickbooks-drive-gmail-claude)
- [Vendor Relationship Tracker: Spot Overpayments and Dying Relationships](https://www.corpusiq.io/blog/vendor-relationship-tracker-overpayments)
---
# Weekly Business Pulse Without a BI Team: GA4 + Shopify + Ads in One Report
URL: https://www.corpusiq.io/blog/weekly-business-pulse-ga4-shopify-ads
Published: 2026-07-20
Category: use-case
Connector: shopify
Build a Monday business pulse from GA4, Shopify, Google Ads, and Meta without a BI team. CorpusIQ gives Claude one cross-source report in seconds.
import { ConnectorCallout } from '@/components/blog/ConnectorCallout'
import { FAQ } from '@/components/blog/FAQ'
Most operators want the same thing on Monday morning: a 10-minute read on how the business did last week. Revenue, traffic, ad performance, any anomaly worth flagging. Not a dashboard that takes two hours to build and updates on a three-day lag. A real-time snapshot across sources.
Connecting GA4, Shopify, Google Ads, and Meta Ads to Claude through CorpusIQ makes this a two-minute exercise. One prompt, four data sources, one cohesive summary.
## Why most "weekly pulse" efforts fail
Three failure modes repeat.
Dashboards that rot. Looker Studio, Triple Whale, Grow, dozens of options. They work until the data source moves, a metric name changes, or nobody updates them. Three months in, the team stops trusting the numbers.
Manual reports that take hours. Pull GA4 into a Sheet, pull Shopify into a Sheet, pull each ads account, merge. Every week. The person who builds it burns out. The dashboard dies with them.
Executive summary emails that are directional at best. Top-line numbers with no context. Revenue was up 12% but nobody can tell you if that was from ads, email, organic, or a one-time customer.
The underlying problem is that weekly reporting requires cross-source synthesis and most small teams cannot afford a data analyst to do it weekly.
## The Claude-assisted pulse
Connect four sources, save one prompt. Rerun it every Monday.
GA4 gives traffic, conversion events, acquisition sources. Shopify gives revenue, orders, AOV, returns. Google Ads gives paid search performance and cost. Meta Ads gives paid social performance and cost.
Ask Claude: "Give me a weekly business pulse for last week versus the prior week. Include revenue and order counts from Shopify. Traffic and conversion rates from GA4, broken down by source. Ad spend and ROAS from Google Ads and Meta. Highlight anything that moved more than 20% in either direction. Keep it to a one-page summary."
Claude queries each connector in parallel and returns a unified report. Changes week over week are front and center. Anomalies are flagged. The CEO gets the Monday read before their first coffee.
## What makes the synthesis valuable
Individually, each connector answers a narrow question. Together, they answer the actual business question.
Revenue was up 15%. Was it more traffic or higher conversion? GA4 says traffic was up 8%, Shopify says conversion was up 6%. Good.
Ad spend was up 30%. Was that deliberate? Google Ads shows new campaigns launched mid-week, Meta spend was flat. Aligned with the plan.
Organic traffic dropped 20%. Any correlated site change? Ask Claude to check if any pages had anomalous traffic changes, then compare to any recent site deploys.
Each question takes seconds. None of them require a dashboard.
## The weekly pulse prompt
Save this. Modify for your business. Rerun weekly.
```
Weekly business pulse for [YOUR BUSINESS], week of [DATE].
Compare last week to the prior week. Include:
1. Revenue and order count from Shopify. AOV. Refund rate.
2. Total sessions, conversion rate, and acquisition source mix from GA4.
3. Ad spend, ROAS, and CPA from Google Ads and Meta Ads, combined and per platform.
4. Any metric that moved more than 20% week-over-week, with a hypothesis about the cause.
5. Top 3 products by revenue, and any that dropped out of last week's top 10.
Format as a one-page summary with short paragraphs, not bullet points. Start with the single most important number.
```
## How to set it up
1. Sign up at [corpusiq.io](https://www.corpusiq.io). Solo $29.95/month covers all four connectors plus 19 more.
2. Connect Shopify, Google Workspace (for GA4), Google Ads, and Facebook Marketing. Each is a standard OAuth flow.
3. Add the CorpusIQ MCP server to Claude.
4. Save the prompt above. Rerun every Monday morning.
## What makes this different from a BI tool
Three things.
Zero maintenance. No dashboards to rebuild when Shopify changes a field name. No ETL jobs to monitor. The prompt is the report.
Cross-source by default. BI tools require you to pre-join data sources. Claude asks all four in the same turn and synthesizes on the fly.
Conversational depth. The report surfaces a number that looks wrong. You ask a follow-up. Claude goes deeper. No ticket to a BI team, no wait, no new dashboard.
The tradeoff: this is not a presentation-grade report with charts. If you need shareable visuals, you still need a dashboard tool. For internal decision-making, the conversation beats the chart.
## See also
- [Connect Shopify to ChatGPT: Daily Store Intelligence Without Dashboards](https://www.corpusiq.io/blog/connect-shopify-chatgpt-daily-store-intelligence)
- [Ad Spend Truth Report: True ROAS Across Google, Meta, and TikTok](https://www.corpusiq.io/blog/ad-spend-truth-report-true-roas)
- [Connect Google Analytics 4 to Claude: Stop Digging, Start Asking](https://www.corpusiq.io/blog/connect-ga4-claude-stop-digging)
---
# Scope Drift Detection: Stop Silent Revenue Leakage
URL: https://www.corpusiq.io/blog/scope-drift-detection-revenue-leakage
Published: 2026-07-20
Category: use-case
Connector: google-drive
Work that creeps beyond scope goes unbilled. Use CorpusIQ to catch scope drift by comparing SOWs in Drive against email and QuickBooks invoices.
import { ConnectorCallout } from '@/components/blog/ConnectorCallout'
import { FAQ } from '@/components/blog/FAQ'
Every services business leaks revenue. The client asks for "one small thing," the account manager says yes, the team does the work, nobody bills for it. Repeat across 30 clients and the leak is real money. The industry benchmark is 5-15% of revenue, which on a $5M agency is $250k-$750k a year.
The problem is detection. Scope drift does not show up on a dashboard. It is buried in email threads, hidden in delivered-but-not-invoiced work, and scattered across SOWs nobody has reread in six months.
Connecting Google Drive, Gmail, and QuickBooks to Claude through CorpusIQ turns scope drift detection from a theoretical good practice into a monthly audit that actually runs.
## Why scope drift is invisible
Three reasons it hides.
First, the leak happens in writing, but the writing is distributed. The SOW is in Drive. The "quick favor" request is in email. The delivery happened on a Zoom call. The invoice that should have included the extra work is in QuickBooks, and the extra line never got added. No single system contains the evidence.
Second, scope creep feels like good client service. Saying yes builds goodwill. Billing for every micro-request feels petty. Account managers who want repeat business default to absorbing the work.
Third, auditing for it is tedious. Reading every SOW, scrolling every email thread, cross-referencing invoices. Nobody has time. So it never gets done.
## The scope drift audit, reframed
The audit logic is simple. For each active client:
1. What does the SOW commit to?
2. What has the client asked for in email since the SOW was signed?
3. What has actually been billed in QuickBooks?
4. Where is the gap between promise, request, and invoice?
Doing this manually for one client takes an hour. For twenty clients, nobody does it.
Doing it with Claude takes five minutes per client.
## The prompt
For each active client, ask Claude:
```
Pull the most recent SOW for [CLIENT NAME] from Drive. Summarize the scope.
Then search Gmail for the last 90 days of emails with anyone from [CLIENT DOMAIN].
List any requests, commitments, or deliverables mentioned that are not explicitly
in the SOW.
Then pull the last 6 months of invoices for [CLIENT NAME] from QuickBooks.
Summarize what was billed and the total.
Compare. Flag: (1) work mentioned in email that does not match the SOW,
(2) work mentioned in email that has no corresponding line item in any invoice,
(3) any pattern suggesting recurring out-of-scope work.
```
Claude synthesizes the three sources. The output is a list of specific items to investigate: "Email dated May 3 from client requesting additional landing page review, not referenced in SOW, no invoice line item found. Possible uncompensated scope expansion."
Act on each one. Either bill for it, document it as a relationship investment, or build a change order template and use it going forward.
## What to do with the findings
Three outcomes for any flagged item.
Bill for it retroactively. Some clients are fine with a "we forgot to invoice this" conversation. Most are, if the work was genuinely delivered and the amount is reasonable.
Document it as an intentional investment. If you chose not to bill as a goodwill gesture, at least know you did. Unseen investments compound into resentment.
Change the process upstream. Repeated flags for the same client or the same type of work are a process signal. Build a change order requirement into your engagement.
## How to set it up
1. Sign up at [corpusiq.io](https://www.corpusiq.io). Solo $29.95/month covers all three connectors.
2. Connect Google Drive (for SOWs), Gmail (for client emails), QuickBooks Online (for invoices).
3. Add CorpusIQ MCP to Claude.
4. Run the audit prompt client by client, monthly or quarterly.
## Running it at scale
For agencies with 30+ clients, the audit becomes a time commitment even with Claude's speed. Two strategies.
Quarterly rather than monthly. The leak rate does not change much month to month. A thorough quarterly audit catches most of what a monthly audit would and takes 25% of the time.
Tiered prioritization. Run the audit in full on your top 20% of clients by revenue. Run a spot-check (SOW versus invoice only, no email scan) on the rest. The revenue leak is Pareto-distributed; the top clients account for most of it.
## What this does not catch
Some scope leakage is invisible even to this audit.
Uncompensated time in internal meetings about client work. Not captured in email or invoices.
Over-servicing that stays within scope. If the SOW allows for "reasonable revisions" and your team does 14 revisions, that is scope absorbed, not scope drift.
Goodwill work done in person or on calls with no written record. Nothing for Claude to query.
The audit catches the writing-based leakage. Which is most of it, at most services businesses.
## See also
- [Services Command Center: Client Health, Scope Drift, and Project Profitability](https://www.corpusiq.io/blog/scope-drift-detection-revenue-leakage)
- [Contract Intelligence Suite: Never Miss a Renewal Again](https://www.corpusiq.io/blog/contract-intelligence-never-miss-renewal)
- [Connect Google Drive to Claude: Turn Every Doc into Answerable Data](https://www.corpusiq.io/blog/connect-google-drive-claude-answerable-data)
---
# The Five Types of MCP Server, Explained
URL: https://www.corpusiq.io/blog/five-types-of-mcp-server
Published: 2026-07-14
Category: deep-dive
A buyer's field guide to the five classes of MCP server: proxy, workflow, unified-API, warehouse-first, and intelligence layer. How to tell them apart.
The Model Context Protocol has become the standard way AI assistants reach outside their own context. ChatGPT, Claude, and Perplexity can all discover and call tools exposed by an MCP server. That standardization is good for everyone. It also created a labeling problem.
Every product in this space now describes itself as an "MCP server," and the phrase has stopped carrying useful information. A server that forwards raw tool calls to 500 apps and a server that reads four systems and validates the numbers between them are both, technically, MCP servers. They solve different problems, fail in different ways, and suit different buyers. Calling them the same thing helps no one who is trying to choose.
This is a field guide to the five classes of MCP server. The point is not to rank them. The point is to give you a vocabulary precise enough that you can tell what you are actually looking at, because the marketing copy usually will not.
## Why the classification matters
The number that most MCP products lead with is integration count. "Connect to 3,000 apps." "500 tools, one endpoint." Breadth is easy to measure and easy to compare, so it dominates the pitch.
Breadth tells you almost nothing about whether a system can answer the question you have. A server connected to 7,000 apps that hands the model raw tool access will still return a wrong revenue figure if two of those apps disagree and nothing reconciles them. A server connected to a dozen systems that validates every number can return an answer you can take to your accountant.
What separates the classes is what the server does with a request. That is the axis worth understanding before you look at any integration count.
## Class one: proxy servers
A proxy MCP server is a universal adapter. It authenticates to many third-party apps and exposes each app's operations as MCP tools, then forwards whatever the AI assistant decides to call. The model picks the tool; the proxy passes the call through and returns the raw result.
Products in this class include Composio and its Rube universal server, Zapier's MCP endpoint, Arcade, Smithery, and Klavis. They compete primarily on breadth and on developer experience: how many apps, how fast to wire up, how good the tool schemas are.
What proxy servers are good at: reach and flexibility. If you are a developer building an agent and you want the widest possible set of tools behind one connection, a proxy is the shortest path. Raw tool access is exactly the right primitive when the calling application is going to supply its own logic on top.
The trade-off is that a proxy does not reason. It has no view of whether two tools returned consistent numbers, no canonical definition of a business metric, no citation trail. Correctness is delegated entirely to the model and to whoever built the agent. For a developer that is a feature. For a business operator who just wants the right answer, it moves all of the hard work downstream to the least controlled part of the stack.
## Class two: workflow servers
A workflow MCP server exposes automations as tools. Underneath, it is an event-and-action engine: when something happens in one system, do something in another. The MCP layer lets an AI assistant trigger or assemble those automations.
Pipedream, n8n, and Make sit here. These are mature automation platforms that added an MCP surface to an existing engine. Their center of gravity is moving data and firing actions between systems in sequence.
What workflow servers are good at: doing things. Create the record, send the message, sync the row, kick off the multi-step process. If your need is action and orchestration, this class is purpose-built for it, with years of connector work behind it.
The trade-off is that a workflow engine is directional and stateful by design. It is built to change things, which is powerful and also the source of its risk. And it is not built to answer an unanticipated cross-system question on demand. Asking a workflow platform "are we profitable this month" is asking a tool designed to move data to instead reason over it, which is not what it does.
## Class three: unified-API servers
A unified-API MCP server normalizes several vendors in one category behind a single schema. Instead of a separate interface for HubSpot, Salesforce, and Pipedrive, you get one "CRM" interface that maps to all of them. The MCP layer exposes that normalized model.
This class comes out of the unified-API tradition, where the value is writing to one abstraction and reaching many providers. It is strongest for software teams building a product that must integrate with whatever CRM, accounting tool, or ticketing system a customer happens to use.
What unified-API servers are good at: portability across vendors in a category. One integration effort, many providers supported, a stable schema that absorbs the differences.
The trade-off is that normalization flattens. To present one schema across many tools, the abstraction has to drop or generalize the fields that do not map cleanly, and those edge fields are often exactly what a specific business question needs. Unified APIs are excellent plumbing for product builders and a poor fit when the answer depends on a provider-specific detail the common schema smoothed away.
## Class four: warehouse-first servers
A warehouse-first MCP server syncs data out of your source systems into a central store it controls, then answers questions against that copy. The MCP layer queries the warehouse rather than the live systems.
Peliqan is the clearest example, with an approach built around a hosted warehouse and comparison content that frames the category around it. The warehouse-first model inherits decades of proven data-engineering practice: a central, queryable store is a well-understood foundation.
What warehouse-first servers are good at: complex historical analysis over large volumes, joins that benefit from everything sitting in one place, and query performance on data that has already been consolidated.
The trade-off is the copy itself. Data in the warehouse is as fresh as the last sync, not the live system. The warehouse becomes a second store of your business data that has to be secured, governed, and reasoned about for retention and residency. For a buyer whose whole reason for caution is not wanting another copy of sensitive data to exist, warehouse-first sits in direct tension with that goal.
## Class five: intelligence layers
An intelligence layer reads multiple live sources on demand, validates the numbers across them, and returns a source-cited answer to the requesting AI client. It does not synchronize records between source systems or maintain a warehouse copy. It treats the multi-source read as the primary operation.
The class is defined by a specific set of components working together: query understanding that turns a sentence into a data plan, a skills router that knows which systems a given question needs, read-only connectors, a metrics registry so "revenue this month" means the same thing every time, a validation engine that surfaces disagreement between sources instead of hiding it, and a citation engine that ties every number back to the record it came from. CorpusIQ is built this way.
What an intelligence layer is good at: cross-system questions where the number has to be correct and checkable. "What did we spend on ads, what did the analytics attribute, and what did we actually collect" is three systems that rarely agree, and the value is in surfacing the gap with citations rather than averaging it away.
The trade-off is honest to state. An intelligence layer is read-first and answer-first for connected vendor systems, so it is not a general workflow-automation engine. CorpusIQ read-only external-source retrieval tools do not write back to vendor systems. CorpusIQ control-plane tools can perform separately named and annotated actions, including changes to user-declared facts, decisions, metric specifications, source manifests, and stored connection state. Connector tools cannot write to a connected system. If you need broad vendor-side automation, this is not that class. If you need to trust what the AI tells you about your systems, it is.
## The five classes side by side
| Class | What it does with a request | Best for | Main trade-off |
| --- | --- | --- | --- |
| Proxy | Forwards raw tool calls to many apps | Developers wanting broad tool access | No reasoning or validation; correctness is downstream |
| Workflow | Triggers event-driven automations | Moving data and taking actions between systems | Directional and write-oriented, not built to answer questions |
| Unified-API | Maps many vendors to one schema | Product teams integrating a whole category | Normalization drops provider-specific detail |
| Warehouse-first | Queries a synced central copy | Historical analysis over consolidated data | Creates a second copy; freshness limited by sync |
| Intelligence layer | Reads live sources, validates, cites | Cross-system answers that must be correct | Read-first; any tool that accepts writes must be separately named and carry safety annotations |
A table like this is a simplification, and the boundaries are not always clean. Some products blur two classes on purpose. But most systems have a center of gravity, and naming it is more useful than counting integrations.
## How to choose
Start from what you actually need the system to produce.
If you need the AI to take actions in your tools, create records, send messages, run multi-step processes, you are shopping for a workflow server, and you should weigh the write-access risk deliberately.
If you are a developer who wants the widest raw tool surface behind one connection and you will supply your own logic, a proxy server is the efficient choice.
If you are building a product that has to integrate with every vendor in a category, a unified-API server is the right abstraction.
If you run heavy historical analysis and are comfortable maintaining a governed copy of your data, warehouse-first earns its keep.
If your problem is that you ask questions spanning several systems and you need the answer to be right and verifiable, without standing up a data pipeline or trusting a raw model call, an intelligence layer is the class built for that job. The test is simple: ask whether the system can show you where a number came from and what it did when two sources disagreed. If it can, you are looking at an intelligence layer. If the answer is "the model figures that out," you are looking at something else.
## Where this leaves the category
MCP solved the connection problem. Any compliant server is reachable by the major assistants, and that is settled. The open question is no longer how to connect. It is what happens after the connection: whether the thing on the other end hands you tools, moves your data, holds a copy of it, or reasons over the live sources and stands behind the answer.
Those are five different products wearing one label. The next time you read "MCP server" in a pitch, the useful question is not how many apps it reaches. It is which of these five it actually is.
CorpusIQ is an intelligence layer: retrieval-only external-source connector tools with separately named and annotated CorpusIQ control-plane tools, scoped customer-data retention, cross-source validation, and citations on every answer, connected to your assistants through MCP. If that is the class your question needs, that is the ground we are built to stand on.
---
# Close the Books 50% Faster: QuickBooks + Drive + Gmail Workflow with Claude
URL: https://www.corpusiq.io/blog/close-books-faster-quickbooks-drive-gmail-claude
Published: 2026-07-13
Category: use-case
Connector: quickbooks
Month-end close is 80% translation work. Learn how to cut close time in half by connecting QuickBooks, Google Drive, and Gmail to Claude through CorpusIQ.
import { ConnectorCallout } from "@/components/blog/ConnectorCallout";
import { FAQ } from "@/components/blog/FAQ";
Month-end close is the most consistent time sink in finance operations. A small-company controller spends three to five business days a month closing the books. Most of that time is not analysis. It is chasing down expense categorizations, matching vendor invoices to payments, and reconciling what happened in the operating business with what the books say.
Connecting QuickBooks, Google Drive, and Gmail to Claude through CorpusIQ collapses the translation work. The close does not get shorter because Claude does accounting. It gets shorter because Claude gathers context in seconds that used to take hours.
## The current close, honestly audited
Walk through a typical monthly close.
Day one: run the P&L and balance sheet. Scan for anything that looks wrong. Export to a spreadsheet.
Day two: chase down miscategorized transactions. Email the ops team asking what a $4,800 Amazon charge was. Open Drive to find the vendor contract that governs a specific recurring expense. Scroll through invoices in Dropbox or Drive to confirm what got paid and what did not.
Day three: reconcile accounts receivable. Look up customer-specific payment terms. Cross-reference email threads where a customer promised to pay "this week." Update the cash forecast.
Day four: build the monthly management report. Pull P&L against budget, AR aging, cash position, key KPIs. Reformat into a template.
Day five: review with the CEO, answer follow-up questions, often by going back to QuickBooks for a different slice of the same data.
The pattern is clear. About 70% of the work is moving information from one place to another. 30% is actual judgment.
## The Claude-assisted close
Same five days, different distribution.
Before close day, run Claude against QuickBooks for a first pass. "Flag any transactions over $2,000 this month that look miscategorized based on vendor name." "List AP with amounts significantly different from the same vendor's historical average." "Which customer balances moved by more than $10,000 this month?"
During close, ask Claude questions as they come up. "What does the contract with Acme say about their payment terms?" Claude queries Drive for the Acme contract, pulls the clause. "What did the ops team say about the $4,800 Amazon charge last month?" Claude searches Gmail for relevant threads.
For management reporting, ask Claude to build the summary directly. "Compare the P&L for this month to the trailing six-month average. Highlight any line that moved more than 20%." "Give me an AR aging report grouped by customer, with any past-due balances flagged."
The close does not take zero days. It takes half as many.
## What changes in practice
Three things change.
First, context is instant. The controller stops waiting on email replies to categorize transactions. If there is any written record (Slack, email, Drive doc), Claude surfaces it in seconds.
Second, the surface area of questions the CEO can ask expands. If a follow-up takes five seconds instead of twenty minutes, more questions get asked. That is the real productivity gain: more iterations per close, which means better understanding, which means better decisions.
Third, the management report becomes a conversation, not a deliverable. The controller and CEO sit down, ask Claude questions, get answers. No one is scrolling through a 12-page PDF.
## How to set it up
Three connectors, one afternoon.
1. Sign up for CorpusIQ at [corpusiq.io](https://www.corpusiq.io). Solo $29.95/month covers QuickBooks, Drive, and Gmail plus the other 40+ connectors.
2. Connect QuickBooks Online via OAuth. Read-only.
3. Connect Google Workspace. Grant Drive and Gmail read-only scopes.
4. Add the CorpusIQ MCP server to Claude.
Test by asking Claude a first question that spans all three: "What is the AR balance for Acme Corp, what does their contract in Drive say about payment terms, and what did the last email thread discuss about the most recent invoice?"
## Sample prompts for a real close
- "Give me a P&L for the month versus the same month last year. Highlight any line item more than 15% different."
- "List every transaction over $1,000 this month that is not categorized."
- "Show me customers with balances over $10,000. For each, summarize the most recent email thread about their payments."
- "Find the master services agreement with Acme in Drive. What does it say about late payment penalties?"
- "Which vendors have we paid more than $5,000 this month and what do the corresponding invoices in Drive confirm we paid for?"
- "Give me a summary of close-relevant emails from the last 30 days: anything mentioning invoice disputes, payment delays, or contract changes."
- "Compare cash position this month to 3 months ago. What are the top 5 line items driving the difference?"
## What stays manual
Claude does not do these. Nor should it.
Judgment on classification when the vendor relationship is genuinely new or unusual. A human has to decide whether that software subscription is marketing or engineering.
Sign-off on the financial statements. Claude produces drafts; a controller or CFO signs.
Tax and compliance filings. Claude can surface source data but a qualified preparer handles filings.
The point of this workflow is not to remove the human from finance. It is to remove the hours of translation between systems so the human spends their time on the 30% of the close that requires judgment.
## See also
- [Connect QuickBooks to Claude: Close Your Books 4x Faster](https://www.corpusiq.io/blog/connect-quickbooks-claude-close-books-faster)
- [Connect Google Drive to Claude: Turn Every Doc into Answerable Data](https://www.corpusiq.io/blog/connect-google-drive-claude-answerable-data)
---
# Connect PostgreSQL to Claude: Natural Language Queries on Production Data
URL: https://www.corpusiq.io/blog/connect-postgresql-claude-natural-language-queries
Published: 2026-07-09
Category: connector-guide
Connector: postgresql
SQL skills are a bottleneck at most companies. Connect PostgreSQL to Claude with CorpusIQ and let anyone query production data safely in plain English.
import { ConnectorCallout } from "@/components/blog/ConnectorCallout";
import { FAQ } from "@/components/blog/FAQ";
Every company has a production database with the answers. Only two or three people can write the SQL. When the CEO asks a question on Tuesday, it gets added to the engineering backlog and answered the following Monday, if at all.
Connecting PostgreSQL to Claude through CorpusIQ breaks that bottleneck safely. Claude converts questions to read-only SQL, runs them, and returns the answer. No one writes to production. Nobody bypasses the DBA.
## The problem with the current workflow
Data access is a classic dependency problem. Product, marketing, finance, and ops all have questions. Engineering has the SQL skills. The queue grows faster than it shrinks. Non-technical stakeholders either get stale data or build workarounds that drift from source of truth.
The usual fix is a BI layer on top. That works for the 20 questions you expect. It fails for the 200 questions you do not.
## How CorpusIQ solves it
CorpusIQ connects PostgreSQL to Claude through MCP. Claude can list tables, describe schemas, and execute read-only SELECT queries.
SELECT-only enforcement. INSERT, UPDATE, DELETE, DROP, ALTER, and other mutating statements are blocked at the connector level. Claude cannot modify the database, regardless of prompt.
Scoped retention. Direct MCP does not retain raw customer files or full connector response payloads. Operational query text, per-user tool-call metadata, and bounded outcome summaries may remain for up to 30 days.
Standard connection. SSL-encrypted connection using credentials you provide. For private databases, point at a read replica or a bastion.
## What you can actually do
- "How many new users signed up this week, broken down by acquisition channel?"
- "Show me the top 20 accounts by MRR and the percentage that are on annual contracts."
- "Which customers have had more than 3 support tickets in the last 30 days?"
- "What is the median time from signup to first paid action for the last 90 days?"
- "Compare churn rate this quarter to last quarter."
- "Show me users who were active daily in the last week but have not upgraded."
- "What is the distribution of subscription plans across the customer base?"
- "Find orders with a refund amount greater than the order value."
Follow-ups work. Ask about churn, break it down by segment, ask which features correlate with retention.
## Setup in 3 minutes
1. Sign up at [corpusiq.io](https://www.corpusiq.io). Solo $29.95/month includes database connectors.
2. From the dashboard, configure the PostgreSQL connection. Host, port, database, read-only user credentials, SSL.
3. Connect CorpusIQ to Claude.
Critical: always connect with a read-only user. CorpusIQ enforces SELECT-only at the connector, but defense in depth is defense in depth.
## Where this earns its keep
Engineering teams tired of serving data requests benefit immediately. So do data-starved product and marketing teams who are used to waiting. The unlock is not that Claude writes perfect SQL: it is that non-technical operators can finally self-serve.
Pairs well with QuickBooks, Shopify, and HubSpot when you want to join your product database against business data.
## What to watch out for
Three caveats.
Claude can write SQL that is technically correct but semantically wrong if your schema is confusing. Table and column names should be descriptive. If columns are called `flag1` and `flag2`, Claude will struggle in the same way a human analyst would.
Permissions should be restrictive. Create a dedicated read-only user with access only to the tables Claude should touch. Do not grant SELECT on tables containing sensitive data (passwords, payment tokens, PHI) unless that is explicitly the intent.
Query cost matters at scale. A naive SELECT on a billion-row table can take minutes and strain your database. Claude will attempt to write efficient queries, but for large tables, consider creating materialized views or running against a read replica.
## See also
- [Database Query Analyst: Plain English Questions Against Your Internal Data](https://www.corpusiq.io/blog/connect-postgresql-claude-natural-language-queries)
- [Connect Airtable to Claude: Your Ops Base as a Question-Answer Engine](https://www.corpusiq.io/blog/connect-airtable-claude-ops-base-qa-engine)
---
# Connect GoHighLevel to ChatGPT: Agency Operations Without Screen-Switching
URL: https://www.corpusiq.io/blog/connect-gohighlevel-chatgpt-agency-ops
Published: 2026-07-06
Category: connector-guide
Connector: gohighlevel
GoHighLevel packs a lot into one screen. Connect GHL to ChatGPT with CorpusIQ for fast client health, pipeline status, and campaign performance.
import { ConnectorCallout } from "@/components/blog/ConnectorCallout";
import { FAQ } from "@/components/blog/FAQ";
GoHighLevel bundles CRM, email, SMS, funnels, calendars, and calls into one platform. That is also the problem: the UI is dense, and answering a simple question often means jumping between four modules.
Connecting GHL to ChatGPT through CorpusIQ collapses those modules into a conversation. Ask what clients need attention, ask which pipelines are stalling, ask which campaigns are performing. One ask, one answer.
## The problem with the current workflow
GHL is built to be a one-stop shop and a lot goes into it. For the operators running it, the tradeoff is that finding specific answers requires navigating multiple modules. Agencies running sub-accounts for clients have this problem multiplied.
Weekly client reporting is the canonical pain. Each sub-account has its own dashboard, and cross-client comparison is manual.
## How CorpusIQ solves it
CorpusIQ connects GoHighLevel to ChatGPT through MCP. ChatGPT can query contacts, opportunities, pipelines, appointments, conversations, and campaigns.
Read-only. No sending, no changes.
Live. Pulls from GHL API directly.
Revocable instantly.
## What you can actually do
- "Show me opportunities across all pipelines that have not moved stage in 30 days."
- "Which sub-accounts have the highest lead-to-appointment conversion rate?"
- "List contacts added in the last 7 days that have not been assigned to a campaign."
- "Compare SMS versus email response rates for this week's campaigns."
- "Which appointments are scheduled for next week, and who owns each?"
- "Find conversations where the last message is more than 48 hours old and was from the client."
- "Summarize the top 5 sub-accounts by opportunity value this month."
- "List all workflows running and their most recent execution count."
Follow-ups work. Ask about stalled opportunities, drill into a sub-account, ask what the conversations show.
## Setup in 3 minutes
1. Sign up at [corpusiq.io](https://www.corpusiq.io). Solo $29.95/month, includes GHL plus 40+ connectors.
2. Connect GoHighLevel via OAuth.
3. Add CorpusIQ to ChatGPT.
## Where this earns its keep
Agencies running GHL as their operating platform for multiple clients benefit most. Weekly reporting, client QBR prep, and internal ops audits get much faster.
## What to watch out for
GHL's API exposes most but not all of what the UI shows. Some deeper campaign analytics are UI-only. Claude returns what the API returns.
## See also
- [Connect HubSpot to Claude: Pipeline Health in One Conversation](https://www.corpusiq.io/blog/connect-hubspot-claude-pipeline-health)
- [CRM Pipeline Health Monitor](https://www.corpusiq.io/blog/connect-hubspot-claude-pipeline-health)
---
# Connect Airtable to Claude: Your Ops Base as a Question-Answer Engine
URL: https://www.corpusiq.io/blog/connect-airtable-claude-ops-base-qa-engine
Published: 2026-07-02
Category: connector-guide
Connector: airtable
Airtable runs most ops stacks but slows down ad-hoc queries. Connect Airtable to Claude with CorpusIQ to ask, summarize, and analyze across bases.
import { ConnectorCallout } from "@/components/blog/ConnectorCallout";
import { FAQ } from "@/components/blog/FAQ";
Airtable is the unofficial ops database for a large slice of small and mid-sized companies. Customer tracking, project management, content pipelines, applicant tracking. The data is rich. Asking it questions requires building a view, which requires knowing what you want to see in advance.
Connecting Airtable to Claude through CorpusIQ lets you ask the base questions directly. No new view, no formula field. Claude pulls the records and answers.
## The problem with the current workflow
Airtable's filtering and sorting are strong for structured workflows. They are weak for ad-hoc questions. If a sales lead asks "which deals in the pipeline have a probability over 70 percent and have not been contacted in 14 days," you are either building a view or exporting to CSV.
For teams where Airtable is load-bearing, the friction adds up.
## How CorpusIQ solves it
CorpusIQ connects Airtable to Claude through MCP. Claude can list bases, read tables and views, and query records with filters.
Read-only. No record changes, no schema modifications.
Live. Every query hits Airtable directly.
Revocable. Remove from Airtable integration page.
## What you can actually do
- "Show me every record in the Customers base with status 'At Risk' and contract value over $50,000."
- "Which projects in the Projects base are overdue?"
- "List all applicants in the ATS base with status 'Phone Screen' and summarize their notes."
- "Compare the count of deals closed this month versus last month."
- "Find duplicate entries across the Contacts table."
- "Which records in the Content Calendar are scheduled for publication this week?"
- "Summarize every record in the Incidents base from the last 30 days."
- "List the top 10 records by any quantitative field."
Follow-ups chain. Ask about at-risk customers, drill into one, ask what the last contact date was, ask who owns the relationship.
## Setup in 3 minutes
1. Sign up at [corpusiq.io](https://www.corpusiq.io). Solo $29.95/month.
2. Click Connect next to Airtable. Approve access.
3. Add CorpusIQ to Claude.
## Where this earns its keep
Teams that use Airtable as a backbone for ops benefit immediately. Agencies, creative teams, recruiters, any workflow where Airtable replaced spreadsheets.
## What to watch out for
Airtable's API has rate limits per base. High-frequency queries may need throttling, which CorpusIQ handles. Very large bases (100k+ rows) can be slow for wide queries; narrow scope when possible.
## See also
- [Connect HubSpot to Claude: Pipeline Health in One Conversation](https://www.corpusiq.io/blog/connect-hubspot-claude-pipeline-health)
- [Connect PostgreSQL to Claude: Natural Language Queries on Production Data](https://www.corpusiq.io/blog/connect-postgresql-claude-natural-language-queries)
- [Database Query Analyst: Plain English Questions Against Your Internal Data](https://www.corpusiq.io/blog/connect-postgresql-claude-natural-language-queries)
---
# Connect TikTok to ChatGPT: Creator Performance Without the Analytics Tab
URL: https://www.corpusiq.io/blog/connect-tiktok-chatgpt-creator-performance
Published: 2026-06-29
Category: connector-guide
Connector: tiktok
TikTok's analytics tab is thin and painful. Connect TikTok to ChatGPT with CorpusIQ for video performance, follower insight, and cross-video trends.
import { ConnectorCallout } from "@/components/blog/ConnectorCallout";
import { FAQ } from "@/components/blog/FAQ";
TikTok's in-app analytics is built for casual creators. For brands and creators running real content programs, the tab is insufficient. Aggregate metrics across videos, finding what is actually working, comparing this month to last, is a manual chore.
Connecting TikTok to ChatGPT through CorpusIQ lets you ask channel-level questions that TikTok's own UI does not answer cleanly.
## The problem with the current workflow
TikTok's analytics show per-video stats reasonably well. Cross-video analysis is where it falls over. Which theme is driving the most engagement? Which content is converting followers? How does current-quarter performance compare to last quarter?
Most creators and brands default to eyeballing recent videos and hoping pattern recognition is enough. It rarely is.
## How CorpusIQ solves it
CorpusIQ connects TikTok's Creator API to ChatGPT through MCP. ChatGPT can query videos, engagement metrics, follower growth, and aggregate performance.
Read-only. No posting, no editing.
Live data. Pulls from TikTok API directly.
Revocable. Remove from TikTok settings.
## What you can actually do
- "Show me my top 10 TikTok videos by views this month."
- "Compare average engagement rate on videos posted in the morning versus evening."
- "What is my follower growth rate this month versus last month?"
- "Which videos in the last 90 days had the highest completion rate?"
- "List videos with high views but low shares."
- "How has my average watch time changed over the past 6 months?"
- "Which of my recent videos had the biggest spike in follower gain?"
- "Show me videos that used the same sound and compare their performance."
Follow-ups keep going. Ask what is working, pick a top video, ask what drove the performance.
## Setup in 3 minutes
1. Sign up at [corpusiq.io](https://www.corpusiq.io). Solo $29.95/month.
2. Connect TikTok via OAuth. Business or Creator account required.
3. Add CorpusIQ to ChatGPT.
## Where this earns its keep
Brands running TikTok as a real acquisition channel benefit. So do agencies managing multiple creators who need to report on performance without manually screenshotting analytics.
## What to watch out for
TikTok's API exposes less than what the in-app analytics tab shows. Some audience demographics are locked to the app. Claude will work with what the API returns.
## See also
- [Connect YouTube to Claude: Channel Analytics and Viewer Insight on Tap](https://www.corpusiq.io/blog/connect-youtube-claude-channel-analytics)
---
# What Is an AI Intelligence Layer for Business Data?
URL: https://www.corpusiq.io/blog/what-is-ai-intelligence-layer-business-data
Published: 2026-06-27
Category: deep-dive
What the AI intelligence layer is, how it differs from connector platforms, ETL tools, and BI dashboards, and why it is a distinct software category.
An AI intelligence layer is a software system that sits between an AI assistant and multiple business data sources. Its job is to accept a plain-English question, determine which business systems contain the data needed to answer it, read those systems in parallel, validate the numbers for consistency, and return a single source-cited answer.
The term distinguishes this type of system from connector platforms, ETL pipelines, BI tools, and generic MCP servers. Each of those categories does something related but structurally different. Understanding the distinction matters because it determines what kinds of questions a system can answer, how it handles data that lives in multiple places, and what security guarantees it can provide.
## The problem an intelligence layer solves
A typical small or mid-size business runs between 10 and 30 software tools simultaneously. Revenue data lives in Shopify or QuickBooks. Customer communication lives in Gmail and Slack. Ad performance lives in Google Ads, Meta Ads, and Google Analytics 4. Inventory and purchasing lives in a separate system. Each tool knows a lot about one domain and nothing about the others.
This creates a structural problem when someone asks a cross-domain question. "Are we profitable this month?" requires pulling gross revenue from Shopify, cost of goods from QuickBooks, ad spend from Google Ads and Meta, and attribution from GA4. No single tool holds all of it. The operator has to open four or five applications, export data, paste it into a spreadsheet, and do the calculation manually. That process takes 30 to 90 minutes and is prone to error at every step.
An AI assistant connected to a single tool cannot solve this. A ChatGPT plugin for QuickBooks tells you what is in QuickBooks. It cannot tell you what the same period looks like when you fold in Shopify revenue and Meta ad spend.
An intelligence layer solves this by treating the multi-source read as a first-class operation. It receives a question, identifies the relevant sources, reads them in parallel, and reasons across the results before returning an answer.
## What an intelligence layer is not
It helps to define the category by contrast, because the terms in this space overlap and the distinctions are not obvious.
### Connector platforms
A connector platform (Zapier, Make, n8n) moves data between systems when a trigger event fires. A new Shopify order creates a row in a Google Sheet. A new HubSpot contact sends a Slack notification. These platforms are directional and event-driven. They do not answer questions. They do not read multiple sources simultaneously. They automate a workflow between two systems in sequence.
An intelligence layer is not event-driven. It is query-driven. A user asks a question; the system reads the data needed to answer it. Nothing moves between systems. Nothing is triggered. Nothing is written.
### ETL pipelines and data warehouses
ETL (Extract, Transform, Load) systems move data out of source systems, transform it into a standard schema, and load it into a central warehouse like BigQuery or Snowflake. The warehouse then becomes the query target.
This approach works at enterprise scale with dedicated engineering resources. It has several costs: data is stale the moment it is loaded, the schema must be defined before questions are asked, and maintaining the pipeline requires ongoing engineering investment.
An intelligence layer reads from source systems on demand, at query time, and returns scoped results to the requesting AI client. It does not build an intermediary warehouse copy. The source systems remain the systems of record, and questions can be asked without first defining a warehouse schema. Operational logs and any optional indexed-search features follow their published retention lifecycles.
### Business intelligence tools and dashboards
A BI tool (Tableau, Looker, Metabase) connects to a data source, lets an analyst define reports and dashboards in advance, and displays the results in a visual interface. The dashboard answers the questions that were anticipated and built at configuration time.
An intelligence layer does not require anyone to define the questions in advance. A user asks a question in plain English that was never anticipated. The system determines what data to read and how to combine it to produce an answer. The set of answerable questions is not bounded by what a dashboard builder thought to include.
### Generic MCP servers
The Model Context Protocol (MCP) is an open standard for exposing tools to AI assistants. Any program that implements the MCP spec becomes discoverable and callable by ChatGPT, Claude, and Perplexity. A basic MCP server exposes a set of tools; the AI assistant calls them during a conversation.
A generic MCP server is infrastructure. It handles the protocol layer: tool definitions, schemas, and transports. It does not include query understanding, skills routing, metrics normalization, cross-source validation, or source citation. Those are application-layer capabilities that a product has to build on top of the protocol.
An intelligence layer uses MCP as the protocol through which AI assistants connect to it. But the intelligence layer itself is a complete system with capabilities that go beyond what the protocol provides.
## The components of an intelligence layer
A production AI intelligence layer for business data has several distinct functional components. Each one is necessary. Removing any one of them degrades the quality of the answers in a specific, predictable way.
### Query understanding
When a user asks "which of my Klaviyo subscribers have not placed a Shopify order in 90 days but opened an email in the last two weeks," the intelligence layer must parse that sentence into a structured data requirement. It needs to know which systems are involved (Klaviyo and Shopify), what the relevant fields are (subscriber status, order date, email open timestamp), what the time ranges are (90 days and 14 days), and how the two datasets should be joined (by email address or customer identifier).
Query understanding translates natural language into a data retrieval plan. Without this component, every question must be pre-specified, or the system can only answer questions that map to predefined tool calls.
### Skills router
A skills router is a catalog of pre-built expert workflows. Each workflow defines what business question it answers, which data sources it reads, in what order, and how the results should be combined and validated.
When a user asks "what is my ad spend efficiency versus actual revenue," the skills router selects the appropriate workflow, which reads Google Ads, Meta Ads, Shopify, and QuickBooks in a specific sequence. The router selects the workflow automatically based on the question, without the user needing to know which workflow applies or which connectors it reads.
Skills encode business logic: which number counts as "ad spend," whether to use attributed or booked revenue, how to calculate ROAS consistently across all users. This prevents each query from producing a different result based on different implicit assumptions.
### Connector layer
The connector layer handles authentication and data access for each business system. Each connector authenticates through the source system's own OAuth implementation, requests only the permission scopes needed to answer questions, and reads data on demand.
The critical design constraint is that each tool's operation and safety boundary is explicit. Read-only external-source retrieval tools use the documented retrieval scopes and do not write back to vendor systems. The only tools that accept writes are CorpusIQ control-plane tools, which manage your own CorpusIQ configuration and never write to a connected system. The boundary is enforced by source scopes and tool-level policy rather than a blanket whole-product claim.
Read-only external-source retrieval limits the impact of any credential issue. A compromised read-only credential can expose data it had access to, but it cannot take action. The blast radius is bounded.
### Metrics registry
A metrics registry is a canonical set of definitions for business KPIs. It specifies, for example, that "revenue this month" means booked orders with a fulfilled status between the first and last day of the current calendar month, using Shopify as the source of truth for orders and QuickBooks as the source of truth for payments received.
Without a metrics registry, cross-source questions return inconsistent numbers. One run uses gross revenue; another uses net. One counts refunded orders; another does not. The metrics registry ensures that the same question asked twice returns the same answer, and that answers across different sessions follow the same definitions.
### Cross-source validation engine
When data comes from multiple sources, discrepancies are common. Shopify and QuickBooks may show different revenue numbers for the same period because of how refunds, chargebacks, and subscription billing are handled differently in each system. An ad platform may report a conversion that GA4 does not record.
A validation engine compares the numbers returned by different sources for consistency before they reach the final answer. When a discrepancy exists, it surfaces it explicitly rather than hiding it in an average or silently preferring one source over another. The user sees that the two systems disagree and by how much.
### Source citation engine
Every data point in the answer is tagged with the originating record and system. If the answer states that "QuickBooks shows $84,320 in accounts receivable as of June 26," the citation traces that number back to the specific AR aging report that produced it. A user can follow the citation to verify the number directly in the source system.
Source citation converts an AI-generated answer from a claim that requires trust into a result that can be verified. This is the primary mechanism for managing hallucination risk in business contexts. Rather than asking a user to trust the output, the system provides the path back to the authoritative record.
### Audit trail
A production intelligence layer can log query text, tool parameters, and bounded result summaries. CorpusIQ records raw query text and tool parameters in local AUDIT logs; its Azure Log Analytics workspace retains those logs for 30 days. Direct MCP does not retain raw customer files or full connector response payloads.
## Security model
The security architecture of an intelligence layer follows from read-only source access and explicit retention lifecycles for direct retrieval, logs, optional indexes, and compliance records.
Read-only external-source retrieval means the system authenticates as a reader of your tools, not as an operator. It can see your Shopify orders. It cannot create or cancel them. It can read your QuickBooks invoices. It cannot create or pay them. This is enforced at the source system level through OAuth scope.
Scoped customer-data retention separates the direct retrieval path from retained operational state. Direct MCP does not retain raw customer files or full connector response payloads. Operational logs may retain query text, per-user tool-call metadata, and bounded outcome summaries for up to 30 days. Optional indexed search retains embeddings and minimal metadata until connector revocation or account deletion. OAuth and account state persist while the connection is active.
This design has direct compliance implications. Source systems remain authoritative, while CorpusIQ's operational logs, optional index, and account state remain governed data classes. Authorized source context also passes to the selected AI client, whose plan and settings govern conversation handling.
## Real examples of cross-source questions
The distinction between single-source and cross-source questions is what separates an intelligence layer from a connected tool. The following are concrete examples of questions that require reading multiple systems simultaneously.
**Marketing attribution validation:**
"What did we spend on Meta Ads last month, what revenue did GA4 attribute to Meta, and what did we actually collect in Shopify from orders that originated from Meta campaigns?"
This requires reading Meta Ads for spend, GA4 for attributed conversions, and Shopify for booked orders. The three numbers rarely match. The gap between them is the attribution discrepancy that the intelligence layer surfaces.
**Collections and customer communication:**
"Which customers have open invoices in QuickBooks that are more than 30 days past due and have not responded to the last three emails in Gmail?"
This requires reading QuickBooks AR aging and Gmail thread history, joined on customer email address, filtered for the intersection of overdue and unresponsive.
**Inventory and demand planning:**
"What are our top 10 Shopify products by revenue this quarter, and do we have enough inventory on hand to cover 30 more days at the current sell-through rate?"
This requires reading Shopify orders for revenue and sell-through rate, and inventory levels for stock on hand, comparing the two to produce a coverage estimate by product.
**Subscriber and order correlation:**
"Which Klaviyo subscribers who clicked a campaign in the last 14 days have not placed a Shopify order in 90 days?"
This requires a list from Klaviyo (clicked, date range) and a list from Shopify (no order, date range), joined on email address.
None of these questions can be answered by querying one system. None of them can be answered by a pre-defined dashboard unless that exact question was anticipated at build time. All of them require a system that reads multiple sources, understands the join condition between them, and returns a single coherent, cited answer.
## How AI assistants connect to an intelligence layer
AI assistants like ChatGPT, Claude, and Perplexity connect to an intelligence layer through the Model Context Protocol. The intelligence layer registers itself as an MCP server, exposing its capabilities as a set of tools. The AI assistant discovers those tools at the start of a conversation and invokes them when a user question requires business data.
When a user asks a question in ChatGPT, the model determines whether the question requires a business data tool, selects the appropriate tool from the intelligence layer's registered capabilities, invokes it with the relevant parameters, receives the structured response, and incorporates the answer into its reply.
The AI assistant does not touch the source systems directly. It interacts only with the intelligence layer, which handles all source system authentication, data retrieval, and validation. The AI assistant handles language understanding and response generation. The intelligence layer handles data access and cross-source reasoning.
This separation has a security consequence: the AI assistant never sees your OAuth tokens. It never holds credentials to your Shopify store or QuickBooks account. All credential management is handled by the intelligence layer, which maintains a controlled audit trail of every operation.
An intelligence layer built for business data typically also exposes access through direct MCP connections for API-based models and agents, and through embedded interfaces in collaboration tools like Slack. In each case, the connector layer, skills routing, and validation logic are the same. The surface through which the user asks the question changes; the system answering it does not.
## The difference from retrieval-augmented generation
Retrieval-Augmented Generation (RAG) is a common architecture for answering questions from unstructured documents. A RAG system chunks documents, embeds them into a vector database, and at query time retrieves the chunks most semantically similar to the question, then passes them to a language model.
RAG is well suited to unstructured text: policy documents, support tickets, contracts, research papers. It is less well suited to structured business data.
The reason is precision. When a user asks "what was our GA4 sessions count for organic search in May," the correct answer is a specific number from a specific system. A RAG system that indexed exported reports can return a chunk of text that contains a number from around that period, but it cannot guarantee that the number is from the canonical source, the correct date range, or the right segment. It approximates.
An intelligence layer executes a structured query against the live source system. The answer is not retrieved from an index of past exports. It is read from the system that owns that data at query time. For numerical business data where precision matters, this is the correct architecture.
RAG and intelligence layers are complementary. An intelligence layer often includes a document search capability for unstructured content within connected systems (email threads, Drive files, Slack messages). That component may use embedding and retrieval internally. But for structured metrics, the intelligence layer queries the source directly.
## Who uses an AI intelligence layer
An intelligence layer for business data is most useful to operators who need to answer questions that span multiple business systems and who lack the engineering resources or time to build a data warehouse and dashboard infrastructure.
This covers individual business operators running an online store, a service business, or a growing company with a combination of software tools. It covers small teams where the person asking the business question is also the person doing the work, with no analyst function to build and maintain dashboards.
It also covers operators who have some BI infrastructure but encounter questions outside its scope. The quarterly report exists in Tableau. The specific cross-tool question that came up in this morning's meeting is not in any dashboard because no one anticipated it.
An intelligence layer answers the questions that were not anticipated. That is its fundamental value relative to every other data system in the category.
## Summary
An AI intelligence layer for business data is a system with a specific architecture: query understanding, skills routing, read-only connector access, metrics normalization, cross-source validation, source citation, and an audit trail. It connects to AI assistants through MCP and returns source-cited answers to plain-English questions that span multiple business systems.
It is distinct from a connector platform, which automates event-driven workflows between tools. It is distinct from an ETL pipeline, which moves data into a warehouse for pre-defined analysis. It is distinct from a BI dashboard, which displays reports that were anticipated at build time. It is distinct from a generic MCP server, which provides protocol infrastructure without application-layer reasoning.
The defining property is cross-source reasoning: the ability to read multiple systems simultaneously, validate the numbers across them, and return a single coherent, cited answer to a question that no single tool could answer on its own.
## How CorpusIQ implements this
CorpusIQ is one implementation of the category described above. Its Intelligence Layer runs twelve engines across three stages: understanding the question, validating the numbers, and applying the right business analysis with citations attached.
For the engine-by-engine breakdown, including how metric definitions stay consistent across ChatGPT, Claude, and Perplexity, see [the CorpusIQ Intelligence Layer](/intelligence-layer).
---
# Connect YouTube to Claude: Channel Analytics and Viewer Insight on Tap
URL: https://www.corpusiq.io/blog/connect-youtube-claude-channel-analytics
Published: 2026-06-25
Category: connector-guide
Connector: youtube
YouTube Studio is built for creators, not operators. Connect YouTube to Claude with CorpusIQ for fast channel analytics and video performance.
import { ConnectorCallout } from "@/components/blog/ConnectorCallout";
import { FAQ } from "@/components/blog/FAQ";
YouTube Studio is built for full-time creators who live in Studio. For operators running a channel as part of a broader content strategy, it is overbuilt. The answer to "is YouTube earning its keep" requires six clicks and still does not feel complete.
Connecting YouTube to Claude through CorpusIQ gives you channel intelligence in plain English. Ask how videos are performing, which ones drive subscribers, where the channel is actually getting traffic from.
## The problem with the current workflow
YouTube Studio has every metric in the world but is painful to navigate for questions that span videos. Which videos drove the most subscribers? What is the geographic mix of viewers? How do Shorts compare to long-form in retention?
Operators typically default to checking views and call it done. That is a superficial read of a channel and leaves most of the insight untapped.
## How CorpusIQ solves it
CorpusIQ connects YouTube to Claude through MCP. Claude can query channel-level analytics, per-video performance, audience geography, traffic sources, and watch time data.
Read-only. Claude cannot upload, edit, or modify videos or channel settings.
Live. Every query hits the YouTube Data and Analytics APIs in real time.
Revocable. Remove from Google account permissions.
## What you can actually do
- "What were my top 5 videos by watch time last month?"
- "Compare subscriber growth this month versus last month."
- "Which videos drove the most channel subscribers in the last 90 days?"
- "Show me the traffic source breakdown for my channel this quarter."
- "What is my average view duration on Shorts versus long-form videos?"
- "List videos with high impressions but low click-through rate."
- "Compare viewer geography this month to last month."
- "Which videos have the highest audience retention past 50 percent?"
Follow-ups drill in. Ask what is working, pick one video, ask what traffic source drove it.
## Setup in 3 minutes
1. Sign up at [corpusiq.io](https://www.corpusiq.io). Solo $29.95/month.
2. Connect Google Workspace with YouTube scope.
3. Add CorpusIQ to Claude.
## Where this earns its keep
B2B companies using YouTube as a content channel benefit most. The question is rarely "are we viral" and often "is this supporting our funnel." Claude plus CorpusIQ answers that by letting you cross-reference YouTube performance against other data sources.
Combines with GA4 to answer "which YouTube videos drove site traffic that converted."
## What to watch out for
YouTube Analytics sampling applies on very wide queries for large channels. Data freshness lags by 24-48 hours for most metrics.
## See also
- [Connect Google Analytics 4 to Claude: Stop Digging, Start Asking](https://www.corpusiq.io/blog/connect-ga4-claude-stop-digging)
- [Connect TikTok to ChatGPT: Creator Performance Without the Analytics Tab](https://www.corpusiq.io/blog/connect-tiktok-chatgpt-creator-performance)
---
# Connect Dropbox to Claude: Contract Intelligence Across Every Folder
URL: https://www.corpusiq.io/blog/connect-dropbox-claude-contract-intelligence
Published: 2026-06-22
Category: connector-guide
Connector: dropbox
Dropbox is where contracts actually live at most small businesses. Connect Dropbox to Claude with CorpusIQ to audit, search, and extract every agreement.
import { ConnectorCallout } from "@/components/blog/ConnectorCallout";
import { FAQ } from "@/components/blog/FAQ";
For most SMBs that have been around for more than five years, Dropbox is where the contracts live. MSAs, NDAs, vendor agreements, client contracts, insurance certificates. Organized by someone who left three years ago, using a naming convention nobody else follows.
Connecting Dropbox to Claude through CorpusIQ turns that archive into something you can actually query. Find every contract with a specific clause, list renewals coming up, summarize what a supplier agreement actually commits you to.
## The problem with the current workflow
Dropbox's search finds files by name and by content inside some file types. It does not answer questions. It does not cross-reference. It does not summarize.
For anyone dealing with contract obligations (which is everyone running a business), the cost is measurable. Missing a renewal window because the agreement was in a folder you forgot about. Signing a second agreement with a vendor because you did not find the first. Agreeing to terms you later find out were already limited by an earlier contract.
## How CorpusIQ solves it
CorpusIQ connects Dropbox to Claude through MCP. Claude can search files by content, list folders, and read docs, PDFs, and standard formats.
Read-only. Claude cannot modify, move, or delete files.
Scoped retention. Queries hit Dropbox API live.
Revocable. Remove from Dropbox's connected apps page.
## What you can actually do
- "Find every contract in Dropbox with an auto-renewal clause and list the renewal dates."
- "Search for NDAs signed in the last 12 months and list the counterparties."
- "Which vendor agreements have a termination-for-convenience clause?"
- "Find insurance certificates and tell me the expiration dates."
- "Search for any contract that mentions a minimum spend commitment."
- "List every client MSA signed in the last 2 years."
- "Find contracts with the word 'exclusivity' in them and summarize what it means in each."
- "Which folders contain files with 'amendment' in the name, and what do they modify?"
Follow-ups chain. Ask about auto-renewals, drill into one, ask what the notice period is, ask what the cost of renewal would be.
## Setup in 3 minutes
1. Sign up at [corpusiq.io](https://www.corpusiq.io). Solo $29.95/month, includes Dropbox plus 40+ connectors.
2. Click Connect next to Dropbox. Approve read-only access.
3. Connect CorpusIQ to Claude.
## Where this earns its keep
Professional services firms, small law firms, real estate operators, and any SMB with a long paper trail benefit most. Contract awareness is undervalued until the day you miss a deadline. Claude plus Dropbox removes that risk for the cost of a connector subscription.
## What to watch out for
Scanned contracts without OCR are unreadable. Large Dropbox tenants with tens of thousands of files may see slower queries for very broad searches. Narrow the folder scope when possible.
## See also
- [Connect Google Drive to Claude: Turn Every Doc into Answerable Data](https://www.corpusiq.io/blog/connect-google-drive-claude-answerable-data)
---
# Connect OneDrive to ChatGPT: Search Company Files in Natural Language
URL: https://www.corpusiq.io/blog/connect-onedrive-chatgpt-search-company-files
Published: 2026-06-18
Category: connector-guide
Connector: onedrive
OneDrive and SharePoint hold the real company memory. Connect OneDrive to ChatGPT with CorpusIQ to search, extract, and summarize Microsoft files.
import { ConnectorCallout } from "@/components/blog/ConnectorCallout";
import { FAQ } from "@/components/blog/FAQ";
For Microsoft-first companies, OneDrive is the filing cabinet. Proposals, contracts, policies, financial models, board decks. Finding anything specific requires remembering which folder someone used five months ago. Search helps, but only if you know the right terms.
Connecting OneDrive to ChatGPT through CorpusIQ turns OneDrive into a queryable knowledge source. Ask what the 2025 budget assumed for payroll, find the latest version of a contract, summarize a deck you only vaguely remember.
## The problem with the current workflow
OneDrive search returns files, not answers. You still open each one. For operators who are on three other apps during the day, that tab-switching is the real cost.
Company knowledge decays fast when nobody can find it. The doc that explains why you chose a specific vendor, the analysis that justified the pricing change, the strategy memo from the last offsite. All of it exists. Almost none of it gets referenced because surfacing it is work.
## How CorpusIQ solves it
CorpusIQ connects OneDrive to ChatGPT through MCP. ChatGPT can search files by content, read Word and Excel and PowerPoint, and extract from PDFs.
Read-only scope. ChatGPT cannot create, modify, or delete.
Live queries. Every request hits Microsoft Graph directly.
Revocable. Remove from Microsoft 365 app permissions.
## What you can actually do
- "Find the 2025 budget spreadsheet and tell me what it assumed for payroll growth."
- "Search OneDrive for any document referencing our data retention policy."
- "Summarize the last three board decks into a one-page brief."
- "Find all contracts with vendors expiring in the next 90 days."
- "Search for files containing our pricing model from the last year."
- "Which documents discuss the acquisition of Acme?"
- "Find Excel files with customer churn analysis and summarize the findings."
- "List all files modified in the last 30 days containing the word 'confidential.'"
Follow-ups drill in. Ask for the budget, ask how payroll grew over time, ask what the assumptions were.
## Setup in 3 minutes
1. Sign up at [corpusiq.io](https://www.corpusiq.io). Solo $29.95/month, includes OneDrive plus 40+ connectors.
2. Click Connect next to Microsoft. Approve read access via Microsoft OAuth.
3. Add CorpusIQ to ChatGPT.
## Where this earns its keep
Microsoft-first enterprises, professional services firms, and any company where OneDrive is the de facto knowledge base benefit most. If the answer to "where is that document" takes more than 30 seconds, Claude or ChatGPT plus CorpusIQ is the fix.
## What to watch out for
Scanned PDFs without OCR are invisible to content search. Tenant-level admin policies may restrict third-party app installation; check with IT first in regulated environments.
## See also
- [Connect Outlook to Claude: Inbox Intelligence for Executives](https://www.corpusiq.io/blog/connect-outlook-claude-inbox-intelligence)
- [Connect Google Drive to Claude: Turn Every Doc into Answerable Data](https://www.corpusiq.io/blog/connect-google-drive-claude-answerable-data)
---
# Connect Microsoft Outlook to Claude: Inbox Intelligence for Executives
URL: https://www.corpusiq.io/blog/connect-outlook-claude-inbox-intelligence
Published: 2026-06-15
Category: connector-guide
Connector: outlook
Outlook is the inbox for most enterprise teams. Connect Outlook to Claude with CorpusIQ to surface priorities, missed threads, and critical actions.
import { ConnectorCallout } from "@/components/blog/ConnectorCallout";
import { FAQ } from "@/components/blog/FAQ";
Most enterprise inboxes are on Microsoft 365. Most executives are drowning in those inboxes. The volume is the problem, the structure is the problem, and Outlook's search is not designed for the "what did I miss this week" question.
Connecting Outlook to Claude through CorpusIQ gives executives a way to triage. Ask what requires a response, ask which threads reference a specific project, ask where the decisions got made this week. Pulls from Exchange Online in real time.
## The problem with the current workflow
Outlook is built for reading one email at a time. It is not built for asking the inbox a question. Search works, rules work, flags work, but none of them answer the question "what do I need to act on today that I have not already seen."
For executives, the cost compounds. Hours of reading email with low information density. Context-switching between Teams, Outlook, and SharePoint. The signal is in the inbox somewhere. Finding it manually burns the productive part of the morning.
## How CorpusIQ solves it
CorpusIQ connects Outlook to Claude through MCP. Claude can search, filter, and summarize across inbox, folders, and specific date ranges.
Read-only. Claude cannot send, reply, archive, or delete.
Scoped retention. Every query hits Microsoft Graph live.
Revocable. Remove from Microsoft 365 app permissions.
## What you can actually do
- "Summarize unresponded emails from the last 7 days that appear to require a decision."
- "Search the last 30 days for any emails mentioning the Acme acquisition."
- "Find every email from my board members in the last 60 days and summarize the top themes."
- "List meetings agreed to via email but not yet on my calendar."
- "Which threads have I been added to in the last 14 days that I have not read?"
- "Search for any legal or compliance notices in my inbox from the last 90 days."
- "Summarize the top 5 longest unread threads."
- "Find emails where my name appears in the body but I am not on the direct To line."
Follow-ups work naturally. Ask what needs response, pick one thread, ask what the full context is, ask what an appropriate reply would cover.
## Setup in 3 minutes
1. Sign up at [corpusiq.io](https://www.corpusiq.io). Solo $29.95/month, covers Outlook plus 40+ connectors.
2. Click Connect next to Microsoft. Approve read access through the Microsoft OAuth flow.
3. Connect CorpusIQ to Claude.
## Where this earns its keep
Enterprise executives, board members, and anyone in a Microsoft-first org gets the most lift. The volume of email at that level is the constraint, and triage is where Claude adds hours back.
Compounds with OneDrive, Calendar, and HubSpot. Ask Claude about a customer across email, files, meetings, and CRM record in one conversation.
## What to watch out for
Microsoft 365 admin policies may restrict OAuth apps. If your IT team blocks third-party apps by default, CorpusIQ will need admin consent before the connection works. Plan for a brief IT conversation.
## See also
- [Connect Gmail to Claude: Recover Critical Emails Before They Cost You](https://www.corpusiq.io/blog/connect-gmail-claude-recover-critical-emails)
- [Missed Critical Email: Recover Time-Sensitive Messages Across Gmail and Outlook](https://www.corpusiq.io/blog/connect-gmail-claude-recover-critical-emails)
- [Connect OneDrive to ChatGPT: Search Company Files in Natural Language](https://www.corpusiq.io/blog/connect-onedrive-chatgpt-search-company-files)
---
# Connect Google Calendar to ChatGPT: Audit Your Time Without Spreadsheets
URL: https://www.corpusiq.io/blog/connect-google-calendar-chatgpt-audit-time
Published: 2026-06-11
Category: connector-guide
Connector: google-calendar
Your calendar is the real record of your week. Connect Google Calendar to ChatGPT with CorpusIQ to audit meeting ROI, recurring waste, and where time goes.
import { ConnectorCallout } from "@/components/blog/ConnectorCallout";
import { FAQ } from "@/components/blog/FAQ";
Most executives cannot tell you how they spent last week. Not because they do not care, because reconstructing it from calendar is tedious. Open Calendar, scroll through every day, manually categorize each meeting. Nobody does this. Everyone suspects they should.
Connecting Google Calendar to ChatGPT through CorpusIQ turns the audit into a conversation. Ask where your time went, which meetings are recurring waste, how much time you spend in recruiting versus customer calls. Data comes from your actual calendar, not your memory.
## The problem with the current workflow
Time audits are high-value and almost never done. The reason is that the tooling is terrible. Spreadsheets require manual entry. Calendar analytics tools require another subscription and another login. Most founders try it once, get a messy answer, and stop.
The miss is costly. Most operators are spending 30-40% of their week in meetings that have no clear ROI. Recurring meetings especially: the Tuesday standup that stopped being useful six months ago, the vendor check-in that should be quarterly not weekly.
## How CorpusIQ solves it
CorpusIQ connects Google Calendar to ChatGPT through MCP. ChatGPT can read events, attendees, durations, and categorize by keywords.
Read-only. ChatGPT cannot create, modify, or decline events.
Live queries. Pulls from Google Calendar API in real time.
Revocable. Disconnect from Google account permissions.
## What you can actually do
- "How many hours did I spend in meetings last week, broken down by meeting type based on title keywords?"
- "List every recurring meeting I have, sorted by total hours this month."
- "Which meetings in the last 30 days had more than 6 attendees, and what was the total cost in person-hours?"
- "Compare my meeting load this month to last month. Am I in more or fewer meetings?"
- "How much time did I spend in 1:1s versus group meetings last quarter?"
- "Show me every meeting over 60 minutes in the last 2 weeks."
- "What percentage of my week was back-to-back with no gaps?"
- "List external meetings (meetings with domains not matching my company) in the last 30 days."
Follow-ups sharpen the audit. Ask how time was spent, drill into the top category, ask which meetings there are recurring, then ask which ones have not had an agenda sent in 90 days.
## Setup in 3 minutes
1. Sign up at [corpusiq.io](https://www.corpusiq.io). Solo $29.95/month, covers Calendar plus 40+ connectors.
2. Click Connect next to Google Workspace. Approve Calendar read access.
3. Add CorpusIQ to ChatGPT.
## Where this earns its keep
Executives, founders, and chiefs of staff benefit most. Anyone whose primary output is decisions gets more leverage from ruthless calendar hygiene than from almost any other intervention.
Stacks well with HubSpot and Gmail. Ask ChatGPT how much time you spent with a specific customer across meetings and email, compared to the revenue they represent.
## What to watch out for
Calendar data is directional, not precise. A 30-minute meeting may actually run 45. A declined invite may still show on the calendar. Use the data for patterns, not for timesheet-level accuracy.
## See also
- [Operations Efficiency Audit: Meeting ROI, Email Overload, and Delegation Readiness](https://www.corpusiq.io/blog/connect-google-calendar-chatgpt-audit-time)
- [Connect Gmail to Claude: Recover Critical Emails Before They Cost You](https://www.corpusiq.io/blog/connect-gmail-claude-recover-critical-emails)
- [Connect HubSpot to Claude: Pipeline Health in One Conversation](https://www.corpusiq.io/blog/connect-hubspot-claude-pipeline-health)
---
# Connect Gmail to Claude: Recover Critical Emails Before They Cost You
URL: https://www.corpusiq.io/blog/connect-gmail-claude-recover-critical-emails
Published: 2026-06-08
Category: connector-guide
Connector: gmail
Emails get buried, misfiled, or trapped in spam. Connect Gmail to Claude with CorpusIQ to surface renewals, payments, and legal notices before they lapse.
import { ConnectorCallout } from "@/components/blog/ConnectorCallout";
import { FAQ } from "@/components/blog/FAQ";
The most expensive email in your inbox is the one you did not read. An auto-renewal notice in spam, a payment reminder buried under newsletters, a legal notice that looked like marketing. Every operator has at least one story about a missed email that cost thousands or broke a relationship.
Connecting Gmail to Claude through CorpusIQ gives you a way to sweep the inbox for things that actually matter. Ask Claude to find recent emails about contracts, payments, legal notices, or anything time-sensitive. Pulls from spam, promotions, archives, every label.
## The problem with the current workflow
Gmail is optimized for replying to threads you are already watching. It is not optimized for audit. Filtering for "everything important I have not handled" is hard. Search works if you know what you are looking for, which is exactly the wrong framing for finding things you have forgotten about.
Most operators compensate by being religious about inbox zero, but that falls apart in busy weeks, and the emails that slip through are disproportionately the ones that matter most. An overdue-invoice reminder looks boring. A contract amendment from legal counsel looks like newsletter clutter.
## How CorpusIQ solves it
CorpusIQ connects Gmail to Claude through MCP. Claude can search messages, read threads, filter by label, and scan folders including spam and promotions.
Read-only scope. Claude cannot send, reply, archive, or delete.
Scoped retention. Every query hits Gmail's API live.
Revocable. Disconnect from Google account permissions.
## What you can actually do
- "Search the last 60 days for any emails mentioning contract renewal, auto-renewal, or expiration."
- "Find messages in spam or promotions that look like legal or compliance notices."
- "List every email from a legal or law firm domain in the last 90 days."
- "Summarize unresponded threads with subject lines containing 'overdue' or 'past due.'"
- "Find all emails where a client complained about a deliverable in the last 30 days."
- "Scan the inbox for vendor renewal notices landing in the next 60 days."
- "List the 10 longest threads I have not responded to this month."
- "Find every email with an attached contract or NDA in the last 6 months."
Follow-ups drill down. Ask for recent legal emails, pick one, ask what it says, ask whether it requires action by a specific date.
## Setup in 3 minutes
1. Sign up at [corpusiq.io](https://www.corpusiq.io). Solo $29.95/month, covers Gmail plus 40+ connectors.
2. Click Connect next to Google Workspace. Approve Gmail read access.
3. Connect to Claude via MCP.
## Where this earns its keep
Founders, operators, and anyone with a high-volume inbox benefit immediately. The time cost of a missed renewal, a missed legal notice, or a missed escalation is disproportionately high relative to the cost of a periodic sweep.
Compounds with HubSpot, QuickBooks, and Drive. Ask Claude what the email thread with a customer said, what the HubSpot record shows about that customer, and what the QuickBooks AR says about their payment history. One question, three systems.
## What to watch out for
Inbox access is sensitive. Claude with Gmail read access can see anything in your inbox including confidential and personal content. Review the scope before connecting. For mixed personal and work accounts, consider a work-only Gmail.
Gmail's API has rate limits. For very large inboxes, broad queries can be slow. Ask narrower questions for faster results.
## See also
- [Missed Critical Email: Recover Time-Sensitive Messages Across Gmail and Outlook](https://www.corpusiq.io/blog/connect-gmail-claude-recover-critical-emails)
- [Connect Google Drive to Claude: Turn Every Doc into Answerable Data](https://www.corpusiq.io/blog/connect-google-drive-claude-answerable-data)
---
# Connect Google Drive to Claude: Turn Every Doc into Answerable Data
URL: https://www.corpusiq.io/blog/connect-google-drive-claude-answerable-data
Published: 2026-06-04
Category: connector-guide
Connector: google-drive
Your real knowledge base is Google Drive. Connect Drive to Claude with CorpusIQ to search, summarize, and extract from every doc, sheet, and PDF.
import { ConnectorCallout } from "@/components/blog/ConnectorCallout";
import { FAQ } from "@/components/blog/FAQ";
Most companies do not have a wiki. They have Google Drive. Every proposal, contract, policy, and SOP lives there. Finding anything specific requires remembering the exact title or rough folder location. Search works, sort of, but only if you guess the right keywords.
Connecting Google Drive to Claude through CorpusIQ turns Drive into something you can ask questions of. What does the vendor contract with Acme say about auto-renewal? Where is the latest version of the pricing sheet? What did the Q3 strategy doc conclude? Plain language, real answers.
## The problem with the current workflow
Drive's search is keyword-based. It returns a list of files that contain the term. It does not read the files for you, summarize across them, or answer questions from their content. You still have to open each one and scan.
For operators, the cost shows up in two places. First, context-switching: opening five docs to answer one question is exhausting. Second, surfacing old knowledge: a doc written 18 months ago might have the exact answer you need today, but you will never find it because you forgot it exists.
## How CorpusIQ solves it
CorpusIQ connects Google Drive to Claude through MCP. Claude can search files by content, list files by folder or type, and read the contents of docs, sheets, and PDFs directly.
Read-only external-source retrieval. Claude cannot create, modify, move, or delete any file.
Direct MCP does not retain raw customer files or full connector response payloads. Every query reads permitted Drive content on demand. Scoped operational logs may persist for up to 30 days; optional indexed search has a separate lifecycle.
Respects Drive permissions. Claude can only read files the connecting user has access to. Sharing permissions in Drive remain the source of truth.
## What you can actually do
- "Find every document that mentions our auto-renewal clause and summarize what each one says."
- "Search all PDFs for our insurance certificates, and list the expiration dates."
- "What does the 2024 strategy doc say about our approach to enterprise customers?"
- "Find the latest version of the pricing sheet and summarize the changes from the previous version."
- "List all contracts signed in the last 12 months."
- "Search Drive for vendor NDAs that have not been countersigned."
- "Summarize the Q3 board prep documents into a one-page brief."
- "Find the onboarding checklist for new hires and show me what it contains."
Follow-ups get sharper. Ask about contracts, drill into one, ask what the termination clause specifies, then ask whether the vendor has a right of first refusal.
## Setup in 3 minutes
1. Sign up at [corpusiq.io](https://www.corpusiq.io). Solo $29.95/month.
2. Click Connect next to Google Workspace. Approve Drive read access.
3. Connect CorpusIQ to Claude.
## Where this earns its keep
Professional services firms, consultancies, and any company with a mature Drive benefit immediately. If your institutional knowledge lives in Drive, Claude becomes the layer that makes it interrogable.
Combines with Gmail and Slack for full-context questions. Ask Claude what the vendor agreement says, what the email thread discussed, and what the Slack conversation decided. Three sources, one answer.
## What to watch out for
Two caveats.
Drive access is broad by scope. When you connect, Claude can see any file the connecting user has access to, including Shared Drives. If your Drive is messy, Claude will see the mess. Consider connecting with a purpose-scoped account if you want to limit exposure.
Image-only PDFs (scanned docs without OCR) are not readable as text. If your Drive has scanned contracts, Claude cannot extract from them until they are OCR'd.
## See also
- [Connect Dropbox to Claude: Contract Intelligence Across Every Folder](https://www.corpusiq.io/blog/connect-dropbox-claude-contract-intelligence)
- [Knowledge Gap Detector: Find Institutional Knowledge That Lives Only in Email and Slack](https://www.corpusiq.io/blog/connect-slack-chatgpt-team-knowledge)
---
# Connect eBay to ChatGPT: Listing Performance and Seller Health on Demand
URL: https://www.corpusiq.io/blog/connect-ebay-chatgpt-listing-performance
Published: 2026-06-01
Category: connector-guide
Connector: ebay
eBay Seller Hub is dense and slow. Connect eBay to ChatGPT with CorpusIQ for fast insight into listings, orders, seller standards, and policy risk.
import { ConnectorCallout } from "@/components/blog/ConnectorCallout";
import { FAQ } from "@/components/blog/FAQ";
eBay's Seller Hub is built for power users running hundreds of listings. For everyone else, it is a maze of tabs and KPIs that shift name every few months. Checking seller standards, recent orders, and underperforming listings takes fifteen minutes if nothing breaks.
Connecting eBay to ChatGPT through CorpusIQ makes those checks seconds long. Ask what the seller health looks like, ask which listings are stalling, ask whether any recent orders have risk flags. ChatGPT pulls from eBay's APIs in real time.
## The problem with the current workflow
eBay sellers manage the business in three places. Seller Hub for orders and listings. Performance dashboard for seller standards. Email and Messages for buyer communication. Cross-referencing them is manual.
The bigger issue is policy and metrics surveillance. eBay's seller standards program is strict. Drift too far on defect rate or late shipment rate and your account tier drops, which means lower visibility and higher fees. Sellers often find out too late because nobody is watching the metrics daily.
## How CorpusIQ solves it
CorpusIQ connects eBay's Seller APIs to ChatGPT through MCP. ChatGPT can query listings, orders, transactions, seller standards, traffic reports, and customer service metrics.
Read-only. ChatGPT cannot list items, modify prices, or respond to buyers.
Scoped retention. Every query hits eBay directly.
Revocable. Disconnect from eBay's Connected Apps page.
## What you can actually do
- "What is my current seller standards program tier, and which metrics are trending toward a downgrade?"
- "Show me listings with more than 100 impressions and zero sales in the last 30 days."
- "What is my current defect rate versus the Top Rated threshold?"
- "List orders from the last 7 days with any delivery issues or open cases."
- "Which listings are getting the most traffic, and which ones have the worst sell-through rate?"
- "Compare this month's GMV to last month, broken down by category."
- "Show me transactions with refunds or returns in the last 30 days."
- "What is my current funds on hold balance and when does it release?"
Follow-ups work. Ask about seller health, drill into the worst metric, then ask which orders are driving that metric.
## Setup in 3 minutes
1. Sign up at [corpusiq.io](https://www.corpusiq.io). Solo $29.95/month, includes eBay plus 40+ connectors.
2. Click Connect next to eBay. Approve the OAuth flow in your eBay account.
3. Add CorpusIQ to ChatGPT.
## Where this earns its keep
Small to mid-size eBay sellers benefit most. You are running a real operation but do not have an analyst watching metrics. ChatGPT plus CorpusIQ becomes your morning health check.
Also useful for resellers running multiple platforms. Pair eBay with Shopify and you can ask ChatGPT about cross-channel SKU performance in one question.
## What to watch out for
eBay's APIs have granular rate limits. Running dozens of broad queries in short succession will trip the limit. For most operators this never matters, but if you are building automations on top, batch queries thoughtfully.
Category-specific policies are not fully exposed through the API. If you sell in restricted categories, Seller Hub is still the source of truth for policy violations.
## See also
- [Connect Shopify to ChatGPT: Daily Store Intelligence Without Dashboards](https://www.corpusiq.io/blog/connect-shopify-chatgpt-daily-store-intelligence)
- [Product Kill or Scale: Rank Every SKU by True Profit](https://www.corpusiq.io/blog/connect-shopify-chatgpt-daily-store-intelligence)
---
# Connect Klaviyo to Claude: Email Revenue Attribution That Actually Works
URL: https://www.corpusiq.io/blog/connect-klaviyo-claude-email-attribution
Published: 2026-05-28
Category: connector-guide
Connector: klaviyo
Klaviyo's dashboards make email look great. Connect Klaviyo to Claude with CorpusIQ to see what actually drives revenue, not just clicks.
import { ConnectorCallout } from "@/components/blog/ConnectorCallout";
import { FAQ } from "@/components/blog/FAQ";
Klaviyo's dashboard tells a happy story. Attributed revenue is big. Open rates look healthy. Every campaign looks like it is earning its keep. The problem is that Klaviyo's attribution window is generous by default, which means some of the revenue showing up under "email" would have happened anyway.
Connecting Klaviyo to Claude through CorpusIQ makes that attribution interrogable. Ask what a specific flow actually added, ask whether a campaign cannibalized a send the next day, ask which segments are dragging overall engagement down. Not a dashboard, a conversation.
## The problem with the current workflow
Klaviyo's reporting is good at one thing: making email look valuable to the person approving the Klaviyo renewal. Campaign reports show opens, clicks, attributed revenue. Flow reports show the same. The top-line number is always directionally correct and always slightly flattering.
For operators trying to actually improve performance, the questions are sharper. What is the incremental revenue from the welcome flow, net of what buyers would have done anyway? Which segment has declining engagement? Are we sending too often to the top 20% of subscribers? These questions require stitching together multiple reports, and Klaviyo's UI does not make that easy.
## How CorpusIQ solves it
CorpusIQ exposes Klaviyo to Claude through MCP. Claude can query campaigns, flows, segments, lists, profiles, and events. That covers campaign performance, flow attribution, segment health, and subscriber lifecycle.
Read-only external-source retrieval. Claude cannot send, pause, edit, or delete anything.
Live data. Every query hits the Klaviyo API. Scoped operational logs may persist for up to 30 days.
Provider-side authorization can be managed in Klaviyo integrations. Disconnecting in CorpusIQ removes its stored connection state and requires reauthorization before reuse.
## What you can actually do
- "Which flows drove the most revenue last month, and what is the revenue per recipient for each?"
- "Show me segments with declining open rates over the last 90 days."
- "Compare abandoned cart flow performance this quarter versus last quarter."
- "Which campaigns had the highest unsubscribe rate, and what was the subject line?"
- "What is my total email revenue this month versus the same month last year?"
- "List the top 10 customers by lifetime email-attributed revenue."
- "Show me every SMS campaign sent in the last 30 days with click-through rate below 2 percent."
- "Which welcome flow variant has the highest conversion rate?"
Follow-ups chain naturally. Ask which flows are working, drill into one, ask which step has the biggest drop-off, then ask what the email at that step looks like.
## Setup in 3 minutes
1. Sign up at [corpusiq.io](https://www.corpusiq.io). Solo $29.95/month, includes Klaviyo plus 40+ connectors.
2. Click Connect next to Klaviyo. Paste your Klaviyo read-only API key.
3. Connect CorpusIQ to Claude via MCP.
## Where this earns its keep
Ecommerce brands using Klaviyo as their primary revenue channel benefit most. The moment email passes 20% of revenue, the questions get real: which flows, which segments, which cadence. Claude plus Klaviyo answers those in seconds.
Compounds with Shopify and Meta Ads. Ask Claude whether email revenue is cannibalizing paid ad attribution, or whether subscribers who came in through Facebook have different LTV than organic subscribers. Cross-source questions are where this stack earns its keep.
## What to watch out for
Two caveats.
Klaviyo's attribution window is a setting. Whatever window you have configured is the lens Claude will answer through. If you attribute last-click for 5 days, that is what Claude reports. Understand your own setting before trusting the numbers.
Segments that are defined with flawed logic will return flawed answers. Claude reports on what segments contain, not whether the segment definition was correct. Audit your segment definitions periodically.
## See also
- [Connect Shopify to ChatGPT: Daily Store Intelligence Without Dashboards](https://www.corpusiq.io/blog/connect-shopify-chatgpt-daily-store-intelligence)
---
# Connect Facebook Ads to ChatGPT: Campaign Analysis Without the Ads Manager
URL: https://www.corpusiq.io/blog/connect-facebook-ads-chatgpt-campaign-analysis
Published: 2026-05-25
Category: connector-guide
Connector: facebook-ads
Meta Ads Manager is slow and cluttered. Connect Facebook Ads to ChatGPT with CorpusIQ for fast campaign, ad set, and creative performance analysis.
import { ConnectorCallout } from "@/components/blog/ConnectorCallout";
import { FAQ } from "@/components/blog/FAQ";
Meta Ads Manager is a reporting tool in name only. It is a buying tool that happens to surface data. Every campaign question takes eight clicks and a filter reset. For operators spending under $100k/month, the UI is an obstacle, not an asset.
Connecting Facebook Ads to ChatGPT through CorpusIQ routes around Ads Manager entirely for analysis. You ask what is working and what is not, ChatGPT pulls from the Meta Marketing API in real time, and you get a ranked answer in seconds.
## The problem with the current workflow
Ads Manager assumes you know exactly which breakdown you want before you open it. If you are looking for underperforming ad sets, you navigate to Ad Sets, filter by date, sort by CPA, scroll. If you want to compare creatives, different screen, different filters. If you want to see how Instagram placements compare to Facebook Feed, different view again.
Operators respond in three predictable ways. Some build a Looker Studio or Shopify Triple Whale dashboard. Those go stale. Some hire an agency and let them own the reporting. That creates dependency. Some just look at the top-line spend and ROAS number and hope for the best. That is where the waste hides.
## How CorpusIQ solves it
CorpusIQ connects Meta's Marketing API to ChatGPT through MCP. ChatGPT gains access to campaigns, ad sets, ads, audiences, placements, and insights across any ad account your Business Manager has access to.
Read-only scope. ChatGPT cannot pause, duplicate, or modify anything. All write actions stay in Meta.
Scoped retention. Direct MCP does not retain raw customer files or full connector response payloads. The Meta API supplies the requested records, and scoped operational logs may persist for up to 30 days.
Provider-side authorization can be managed in Meta Business Manager. Disconnecting in CorpusIQ removes its stored connection state and requires reauthorization before reuse.
## What you can actually do
- "Which ad sets have spent over $500 this month with a ROAS below 1?"
- "Show me my top 10 creatives by CTR in the last 14 days."
- "Compare Instagram Feed versus Facebook Feed performance for the current campaign."
- "List campaigns where CPA has risen more than 30 percent week over week."
- "Which audiences have the highest frequency, and what is their conversion rate?"
- "Show me every ad with a relevance score below average that has spent more than $100."
- "What is my blended CPM across campaigns this month versus last month?"
- "Flag any ads with high impressions but low clicks in the last 7 days."
Follow-ups work naturally. Ask which ad sets are underperforming, then ask what creatives they are running, then ask which placements are eating the budget. One conversation, three levels deep.
## Setup in 3 minutes
1. Sign up at [corpusiq.io](https://www.corpusiq.io). Solo $29.95/month covers Meta Ads plus all 40+ connectors.
2. Click Connect next to Facebook Marketing. Log in to Meta, select your Business Manager and ad accounts, approve read-only.
3. Add the CorpusIQ connector to ChatGPT.
## Where this earns its keep
Direct-to-consumer brands under $20M in revenue get the most lift. You are running enough ads to have real data, not enough budget to staff a dedicated paid media analyst. ChatGPT plus CorpusIQ replaces the weekly audit you never quite have time for.
Stacks well with Shopify, Google Ads, and TikTok. Ask ChatGPT for blended ROAS across channels in one question instead of checking three dashboards.
## What to watch out for
Two caveats.
Attribution in Meta is Meta's attribution. If your pixel setup is noisy or your Conversions API is sending bad events, the numbers ChatGPT returns will reflect that. Fix the tracking before trusting the insights.
iOS 14+ continues to degrade what Meta can report on at the ad level. Claude will return what the API gives back. The ceiling on insight depth is set by Apple, not CorpusIQ.
## See also
- [Connect Google Ads to Claude: Find Wasted Spend in Minutes](https://www.corpusiq.io/blog/connect-google-ads-claude-find-wasted-spend)
- [Connect Shopify to ChatGPT: Daily Store Intelligence Without Dashboards](https://www.corpusiq.io/blog/connect-shopify-chatgpt-daily-store-intelligence)
---
# Connect Google Ads to Claude: Find Wasted Spend in Minutes
URL: https://www.corpusiq.io/blog/connect-google-ads-claude-find-wasted-spend
Published: 2026-05-21
Category: connector-guide
Connector: google-ads
Google Ads has 30 screens to find what wastes money. Connect Google Ads to Claude with CorpusIQ to surface weak keywords, ads, and campaigns fast.
import { ConnectorCallout } from "@/components/blog/ConnectorCallout";
import { FAQ } from "@/components/blog/FAQ";
Every Google Ads account has waste. The question is where. The Google Ads interface makes finding it genuinely difficult: different reports for keywords, search terms, ad variations, assets, and placements, each with their own filters and date ranges. A thorough audit takes half a day. Most operators skip it.
Connecting Google Ads to Claude through CorpusIQ compresses that half-day into a conversation. Ask where the waste is, Claude pulls from the Google Ads API, and the underperformers surface in minutes.
## The problem with the current workflow
Google Ads' UI is built for buyers who live in it. If you are managing a $20k/month spend as one of ten things on your plate, the interface punishes you. Keyword performance here, search terms there, asset reports in a third place, placements in a fourth. Each one is a filter-sort-export ritual.
Agencies fill this gap, but agency reports tend to focus on what the agency wants to highlight. The campaigns that are working get prominent placement. The campaigns that are bleeding money get a footnote. The question "where is my waste" is not the question agency reports are designed to answer.
Doing it yourself requires fluency in the Google Ads Editor or API. Most operators do not have that fluency. The result is wasted spend that stays wasted for months.
## How CorpusIQ solves it
CorpusIQ connects Google Ads to Claude through MCP. Claude can query campaigns, ad groups, keywords, ads, search terms, audiences, assets, and performance metrics across any account or MCC you grant access to.
Read-only access. CorpusIQ requests only the read scope. Claude cannot pause campaigns, adjust bids, modify budgets, or make any structural change to your account. It surfaces what is happening. You decide what to do.
Live queries. Every question hits the Google Ads API in real time. No caching, no delayed reports, no stale data.
Provider-side authorization can be managed from Google account permissions. Disconnecting in CorpusIQ removes its stored connection state and requires reauthorization before reuse.
## What you can actually do
- "Which keywords have spent more than $500 this month with zero conversions?"
- "Show me the top 20 search terms triggering my ads that I have not added as exact match or negatives."
- "List campaigns where CPA is more than 50 percent above the account average."
- "Which ad groups have impressions but no clicks in the last 14 days?"
- "Compare this month's CTR to last month, broken down by campaign."
- "What is my cost per conversion on brand terms versus non-brand terms?"
- "Show me every ad variation with a quality score below 5."
- "Which geo locations are converting below the account average but getting more than 10 percent of the spend?"
Follow-ups are natural. Ask about wasted spend, drill into the top offender, then ask what search terms are driving that spend. Three questions, concrete audit findings.
## Setup in 3 minutes
1. Sign up at [corpusiq.io](https://www.corpusiq.io). Solo $29.95/month includes Google Ads plus all 40+ connectors.
2. Click Connect next to Google Ads. Authorize read-only access. If you manage an MCC, grant MCC access and CorpusIQ can query every child account.
3. Add the CorpusIQ MCP server to Claude.
First question is live.
## Where this earns its keep
The operators who get the most value from this are the ones managing Google Ads themselves or supervising an agency. If you run your own ads, Claude replaces the audit you keep meaning to do. If you work with an agency, Claude gives you an independent view of what is happening in the account so you can ask sharper questions during QBRs.
It stacks well with GA4, Shopify, and Klaviyo. Ask Claude about wasted Google Ads spend, then ask which landing pages those ads sent traffic to, then ask whether any of that traffic converted downstream. The cross-source view is where the real analysis lives.
It also helps with attribution sanity checks. If Google Ads says a campaign drove 50 conversions but Shopify and GA4 tell a different story, you can surface the gap in one question. The discrepancy is the signal.
## What to watch out for
Three honest caveats.
First, Claude will not replace a senior PPC manager for strategy. It is excellent at surfacing problems and asking sharper questions. It is not a substitute for someone who knows when to shift from Manual CPC to Target CPA, or how to structure a PMax campaign correctly.
Second, Performance Max is a black box by Google's design. The API exposes asset-level and campaign-level data, but deep breakdowns by audience or placement inside PMax are limited. Claude returns what Google returns. The ceiling on PMax insight is set by Google, not by CorpusIQ.
Third, attribution is a Google Ads reporting choice. The conversion counts Claude surfaces reflect whatever attribution model you have set. If that model does not match reality, the answers will look off. Configure attribution thoughtfully before auditing.
## See also
- [Connect Google Analytics 4 to Claude: Stop Digging, Start Asking](https://www.corpusiq.io/blog/connect-ga4-claude-stop-digging)
- [Connect Facebook Ads to ChatGPT: Campaign Analysis Without the Ads Manager](https://www.corpusiq.io/blog/connect-facebook-ads-chatgpt-campaign-analysis)
---
# Connect Slack to ChatGPT: Turn Team Chatter into Searchable Knowledge
URL: https://www.corpusiq.io/blog/connect-slack-chatgpt-team-knowledge
Published: 2026-05-18
Category: connector-guide
Connector: slack
Most company knowledge lives in Slack threads nobody can find. Connect Slack to ChatGPT with CorpusIQ to surface decisions buried in DMs and channels.
import { ConnectorCallout } from "@/components/blog/ConnectorCallout";
import { FAQ } from "@/components/blog/FAQ";
Slack is where most operating decisions get made and none of them get filed. The reasoning behind a pricing change, the context on why a vendor was chosen, the thread that explained the weird exception in the contract. All of it lives somewhere in Slack, and Slack's search has a reputation for a reason.
Connecting Slack to ChatGPT through CorpusIQ turns your team's accumulated chatter into something you can actually query. Ask a question, get an answer that cites the actual threads where the decision happened.
## The problem with the current workflow
Every company past 10 people has the same problem. Decisions happen in Slack. Nobody writes them down. Three months later, someone asks "why did we do it this way" and the answer exists, but only if you can find the thread.
Slack's native search is keyword-based and paginates badly. You type a term, you get 200 results, you scroll until your eyes glaze over. If the original message used slightly different words than you remember, you never find it. The information is not lost. It is just not reachable.
Teams respond by creating wikis or Notion pages, but the wikis go stale and nobody updates them. The authoritative version of most operating context stays in Slack whether you like it or not.
## How CorpusIQ solves it
CorpusIQ connects Slack to ChatGPT through MCP. ChatGPT can search messages, read threads, look at channel context, and summarize discussions across the parts of your workspace you have granted it access to.
The connection is read-only. ChatGPT cannot post messages, create channels, add users, or do anything that modifies your Slack workspace. The only action available is reading.
Scoped retention. CorpusIQ does not copy Slack messages into a database. Every search hits the Slack API and returns the result. Messages do not persist in CorpusIQ.
Scoped access. The Slack OAuth install respects workspace policies. You choose which channels the integration can access when you install. Private channels require explicit invitation. DMs are scoped to the installing user.
## What you can actually do
- "Find the thread where we decided to go with the new pricing structure and summarize the key points."
- "What did the engineering team say about the database migration last week? Summarize across the top 3 threads."
- "Search the customer-success channel for any mentions of customer complaints about onboarding in the last 30 days."
- "Who was working on the Q2 marketing plan, and what is the latest status based on the Slack threads?"
- "Find the thread where legal weighed in on the vendor contract and give me the bullet points."
- "What decisions were made in the #leadership channel in the last week?"
- "Summarize the standup threads in #engineering for this week. What is blocked, what is shipping?"
- "Find all Slack messages where someone mentioned the Acme account in the last 60 days."
This turns Slack from a firehose into a knowledge source. The threads were always there. Now you can reach them.
## Setup in 3 minutes
1. Sign up at [corpusiq.io](https://www.corpusiq.io). Solo $29.95/month covers Slack plus all 40+ connectors.
2. Click Connect next to Slack. Choose your workspace. Approve the scopes.
3. Add the CorpusIQ connector to ChatGPT from the ChatGPT connectors directory.
Ask your first question.
## Where this earns its keep
The companies that benefit most are knowledge-work teams between 10 and 200 people. Small enough that key context lives in Slack. Large enough that you cannot scroll back to find it.
It works especially well for new hires. A new employee can ask ChatGPT "what have we decided about X" and get an answer sourced from the actual threads, instead of asking three people in DMs who will each give a slightly different version.
For operators and founders, it compounds with other connectors. Ask ChatGPT about a customer complaint, then ask what the HubSpot record says, then ask what the Gmail thread shows. The Slack context fills in the gaps that formal systems miss.
## What to watch out for
Three things to consider before you connect Slack.
First, permissions. ChatGPT with Slack access can read anything the installing user or the app scope can read. If you install at the workspace level, admins should review the scope list carefully. For most teams, installing with minimum scopes and adding the app to specific channels is the safer path.
Second, sensitive content. Slack channels often contain PII, client details, and confidential strategy. Your organization's data policies apply to what gets surfaced through ChatGPT. If you are in a regulated industry, talk to compliance before connecting.
Third, noise. Slack has a lot of noise. ChatGPT will work with what is there. If your channels are mostly memes and GIFs, search quality will reflect that. Integrations do better with workspaces that use channels thoughtfully.
## See also
- [Knowledge Gap Detector: Find Institutional Knowledge That Lives Only in Email and Slack](https://www.corpusiq.io/blog/connect-slack-chatgpt-team-knowledge)
- [Connect Gmail to Claude: Recover Critical Emails Before They Cost You](https://www.corpusiq.io/blog/connect-gmail-claude-recover-critical-emails)
---
# Connect HubSpot to Claude: Pipeline Health in One Conversation
URL: https://www.corpusiq.io/blog/connect-hubspot-claude-pipeline-health
Published: 2026-05-14
Category: connector-guide
Connector: hubspot
Forecast reviews, deal stalls, rep productivity, contact engagement. Connect HubSpot to Claude with CorpusIQ and get answers without building another dashboard.
import { ConnectorCallout } from "@/components/blog/ConnectorCallout";
import { FAQ } from "@/components/blog/FAQ";
Pipeline review meetings are where data goes to be argued with. The report says one thing, the rep says another, the forecast gets adjusted on vibes. Nobody leaves the room confident. Nobody goes back to HubSpot to check the actual numbers, because pulling them would take another 20 minutes.
Connecting HubSpot to Claude through CorpusIQ lets you ask pipeline questions the same way you ask any other question. In plain English, during the meeting, with the answer coming straight from HubSpot. No report builder, no filtered view, no hunting for the right property.
## The problem with the current workflow
HubSpot is a capable CRM but its reporting model punishes ad-hoc questions. You can build a beautiful dashboard for your top 10 recurring metrics. You cannot, without effort, answer the question someone raises in a Tuesday standup. Custom reports take time. Saved filters get forgotten. Every rep has their own view of their own pipeline.
The predictable result is that pipeline conversations happen on feel. The sales leader knows which deals "feel solid" and which ones "are slipping." The CFO asks for a forecast and gets a number that was manually adjusted three times before anyone agreed on it. Nobody is working from the same data.
The CRM has the answers. The problem is that extracting them requires either a dedicated RevOps person or a level of HubSpot fluency most teams do not have.
## How CorpusIQ solves it
CorpusIQ exposes HubSpot to Claude through MCP. Claude gains the ability to query contacts, companies, deals, tickets, pipeline stages, owners, activities, and any custom properties your team uses. Ask a pipeline question and Claude pulls the answer from HubSpot in real time.
The HubSpot retrieval tools request read scopes and do not modify deals, reassign owners, change stages, or delete contacts. This statement holds across the connector catalog. The only tools that accept writes are CorpusIQ control-plane tools, which manage your own CorpusIQ configuration.
Scoped retention. Direct MCP does not retain raw customer files or full connector response payloads. Every query reads HubSpot on demand; operational logs may persist for up to 30 days.
Provider-side authorization can be managed from HubSpot's integrations panel. Disconnecting in CorpusIQ removes its stored connection state and requires reauthorization before reuse.
## What you can actually do
- "Show me all deals over $25,000 in the proposal stage that have not had activity in 14 days."
- "Which deals closed this quarter, broken down by source, and which sources have the highest win rate?"
- "Give me a pipeline forecast for the next 90 days using weighted stage probabilities."
- "List contacts at companies with more than 100 employees who have opened an email in the last 7 days but have no assigned owner."
- "Which reps have the longest average time from opportunity to close, and what is the average deal size for each?"
- "Show me deals where the close date has been pushed more than twice, grouped by owner."
- "What is the conversion rate from MQL to SQL to Customer over the last 6 months?"
- "Flag any enterprise deals over $50k with no meeting scheduled in the last 21 days."
Follow-ups behave like a conversation. Ask about stalled deals, then narrow to one owner, then ask what the common reason in the notes field is. Claude keeps the context.
## Setup in 3 minutes
1. Create an account at [corpusiq.io](https://www.corpusiq.io). Solo $29.95/month covers HubSpot plus all 40+ connectors.
2. In the dashboard, click Connect next to HubSpot. Authorize read-only access in HubSpot.
3. Connect Claude to CorpusIQ via the MCP setup in your Claude dashboard.
First question is live.
## Where this earns its keep
The teams that benefit most have a HubSpot instance with real data in it and not enough analyst hours to serve everyone who has questions. Sales leaders, RevOps, and founders running their own pipeline all save hours a week.
It compounds with Gmail, Outlook, and Google Calendar. Ask Claude about a stalled deal, then ask what the last email thread said, then ask when the next meeting is. Three connectors, one conversation. CorpusIQ's MCP layer is what makes the cross-source questions work.
It also pays off at forecast time. Instead of the sales leader spending 4 hours building a deck, they ask Claude for the forecast and the stage-by-stage breakdown, then review it with the team. The conversation is about what to do, not about how to build the slide.
## What to watch out for
Two honest caveats.
Data quality matters. If your reps do not update deal stages, the forecast will be wrong. Claude reports what HubSpot contains. It cannot guess what was supposed to be there.
Custom property coverage varies. HubSpot exposes standard properties reliably through the API. Custom properties work but require that your HubSpot admin has kept the property schema clean. Ask Claude to list available properties first if you are unsure what is queryable.
## See also
- [Connect Gmail to Claude: Recover Critical Emails Before They Cost You](https://www.corpusiq.io/blog/connect-gmail-claude-recover-critical-emails)
- [Connect Google Calendar to ChatGPT: Audit Your Time Without Spreadsheets](https://www.corpusiq.io/blog/connect-google-calendar-chatgpt-audit-time)
---
# Connect Google Analytics 4 to Claude: Stop Digging, Start Asking
URL: https://www.corpusiq.io/blog/connect-ga4-claude-stop-digging
Published: 2026-05-11
Category: connector-guide
Connector: ga4
GA4 is powerful and painful. Connect GA4 to Claude with CorpusIQ for traffic, conversion, and funnel answers in plain English, no Explorations.
import { ConnectorCallout } from "@/components/blog/ConnectorCallout";
import { FAQ } from "@/components/blog/FAQ";
GA4 knows the answer. Finding it is the problem. The event-based model is more flexible than Universal Analytics was, but the UI makes you pay for that flexibility in clicks. Explorations, dimension pickers, date ranges, comparison toggles. By the time you have built the report, the question has gone cold.
Connecting GA4 to Claude through CorpusIQ skips the UI entirely. You ask the question, Claude pulls the answer from the GA4 API, and the analysis comes back as a sentence, not a chart you have to interpret.
## The problem with the current workflow
GA4's interface is optimized for analysts who live in the tool. For operators who dip in once a week, the friction is punishing. Every question requires setting up dimensions, metrics, filters, and comparisons from scratch. Save a report, forget where you saved it, rebuild it next Tuesday.
Most small teams respond by either building a Looker Studio dashboard that goes stale or ignoring GA4 entirely and relying on Shopify or Stripe as ground truth. Both are bad outcomes. Looker dashboards rot. Ignoring GA4 means you cannot answer questions about traffic sources, landing page conversion, or funnel drop-off.
The deeper issue is that GA4 rewards people who already know what to look for. If you know to check landing page bounce rate by source for organic traffic over the last 28 days, GA4 will show it to you. If you are just wondering whether last week's blog post drove anything, you are in for a 15-minute click-fest.
## How CorpusIQ solves it
CorpusIQ connects your Google account to Claude through MCP. Once connected, Claude can query the GA4 Data API directly. Traffic reports, conversion events, funnel analysis, audience breakdowns, real-time users, and ecommerce events are all accessible through natural language.
The architecture matches the rest of CorpusIQ:
Read-only external-source retrieval. CorpusIQ requests only the GA4 read scope. Claude cannot modify your property settings, add users, or change data retention.
Direct MCP does not retain raw customer files or full connector response payloads. Claude asks a question and CorpusIQ queries the GA4 API. Scoped operational logs may persist for up to 30 days.
Provider-side authorization can be managed from Google account permissions. Disconnecting in CorpusIQ removes its stored connection state and requires reauthorization before reuse.
## What you can actually do
- "What were my top 10 landing pages by conversion rate last month, and how do they compare to the same page set the month before?"
- "How much organic search traffic did I get last week, broken down by country?"
- "Which blog posts drove the most newsletter signups in the last 30 days?"
- "Show me the funnel from homepage to pricing page to signup for paid traffic only."
- "What is my average session duration by traffic source, and which source has the lowest bounce rate?"
- "Compare this week's traffic to last week. Highlight any sources that are up or down more than 25 percent."
- "Which pages are getting traffic but not converting, ranked by sessions?"
- "What is the real-time user count right now, and which campaign is driving the most active sessions?"
Follow-ups work. Ask for top landing pages, then ask which ones had declining conversion rates, then ask whether the decline correlates with a specific source. Three questions, no dashboard.
## Setup in 3 minutes
1. Sign up at [corpusiq.io](https://www.corpusiq.io). Solo plan $29.95/month, includes GA4 and all 40+ connectors.
2. In the dashboard, click Connect next to Google Workspace. Grant read-only access including Google Analytics. CorpusIQ auto-discovers every GA4 property you have access to.
3. Add the CorpusIQ MCP server to Claude. First question is live.
## Where this earns its keep
The teams that get the most from this are marketing operators and founders who need to answer traffic questions fast but do not have an analyst. If you are already paying an agency for a monthly GA4 report, connecting GA4 to Claude gives you the ability to ask your own questions between reports. You stop waiting for the next deliverable.
It also stacks well with Google Ads, Facebook Marketing, and Shopify. Ask Claude about organic traffic trends, then ask which paid campaigns correlate with the organic lift, then ask how both map to revenue. Cross-source analysis is where CorpusIQ earns its keep.
## What to watch out for
GA4 has known quirks that Claude cannot paper over.
First, GA4 data has a processing delay. The same-day and last-24-hours numbers are not final for 24-48 hours. Claude will return what GA4 returns, which may shift the next day. This is a GA4 limitation, not a CorpusIQ one.
Second, GA4's thresholding applies. If your property has low traffic, some dimensions will be suppressed to protect user privacy. Claude will tell you when a query returns thresholded data. This is correct behavior, not a bug.
Third, GA4 sampling kicks in at scale. Properties over the sampling threshold will return sampled results for wide queries. Claude will surface the sample rate when it is relevant. For precise numbers at high scale, you need BigQuery export, which is outside CorpusIQ's scope.
## See also
- [Connect Google Ads to Claude: Find Wasted Spend in Minutes](https://www.corpusiq.io/blog/connect-google-ads-claude-find-wasted-spend)
---
# Connect Shopify to ChatGPT: Daily Store Intelligence Without Dashboards
URL: https://www.corpusiq.io/blog/connect-shopify-chatgpt-daily-store-intelligence
Published: 2026-05-07
Category: connector-guide
Connector: shopify
Skip the Shopify admin and BI dashboards. Connect Shopify to ChatGPT with CorpusIQ for live revenue, product performance, and customer insight.
import { ConnectorCallout } from "@/components/blog/ConnectorCallout";
import { FAQ } from "@/components/blog/FAQ";
Most Shopify operators have three browser tabs open at all times. The admin panel for orders. A BI tool or spreadsheet for trends. Something like Triple Whale or Northbeam for attribution. By the time you have pulled yesterday's numbers across all three, it is lunch, and the real question is still sitting there unanswered.
Connecting Shopify to ChatGPT through CorpusIQ removes most of that tab-switching. You ask a question, ChatGPT pulls from Shopify in real time, and the answer comes back in plain English. No dashboard to build, no saved view to remember.
## The problem with the current workflow
Shopify's admin is built for running the store, not for thinking about it. Filters are narrow, comparisons are limited, and once you want to slice revenue by something non-obvious, you end up exporting to CSV. Every operator knows this ritual. Export, open Sheets, pivot, lose an hour.
BI tools help but add cost, latency, and a maintenance burden. Most stores under $10M in revenue do not have a dedicated analyst, so the founder or ops lead ends up owning the dashboard. It slowly drifts out of date. Three months in, nobody trusts the numbers.
The underlying issue is that store questions come up at odd hours and require instant answers. What did yesterday close at? Which SKU is spiking? Is that Klaviyo flow pulling its weight? These questions do not justify a 30-minute dashboard build. They justify a 30-second answer.
## How CorpusIQ solves it
CorpusIQ connects Shopify to ChatGPT through the Model Context Protocol. Once connected, ChatGPT can query your store data in real time: orders, products, customers, inventory, fulfillment, refunds, and more.
Three design choices keep this safe and useful:
Read-only access. CorpusIQ asks Shopify for the minimum scopes required to answer questions. ChatGPT cannot create orders, change prices, modify inventory, or touch anything in the store. The only risk surface is what ChatGPT can read.
Scoped retention. Direct MCP does not retain raw customer files or full connector response payloads; operational logs may persist for up to 30 days. When ChatGPT needs data, CorpusIQ fetches it from Shopify and returns a cited answer. ChatGPT conversation handling follows the OpenAI plan and settings you choose.
Disconnecting in CorpusIQ removes its stored connection state and requires reauthorization before reuse. Provider-side authorization remains governed by Shopify and can be managed in Shopify's app settings.
## What you can actually do
Real prompts, live data.
- "What did we do in revenue yesterday compared to the same day last week, and which products drove the difference?"
- "Show me the top 10 products by revenue this month, and how their refund rates compare to the store average."
- "Which customers have spent over $500 lifetime but have not ordered in 90 days?"
- "What is my current inventory position on SKUs that sold more than 50 units last month?"
- "List orders over $300 placed in the last 7 days with any fulfillment issues."
- "Which product variants are converting below 1 percent when viewed, and how much traffic are they getting?"
- "Compare AOV this month to last month. Break it down by new versus repeat customer."
- "Flag any abandoned checkouts over $200 from the last 48 hours."
Each follow-up works naturally. Ask about yesterday's revenue, then drill into the top product, then ask which ad campaign that product's buyers came from. ChatGPT holds the thread.
## Setup in 3 minutes
Standard OAuth flow, no developer work required.
1. Create an account at [corpusiq.io](https://www.corpusiq.io). Solo plan $29.95/month includes Shopify and all 40+ connectors. 30-day free trial.
2. In the CorpusIQ dashboard, click Connect next to Shopify. Enter your store domain, approve read-only access.
3. In ChatGPT, add the CorpusIQ connector from the ChatGPT connectors directory. Log in with your CorpusIQ credentials.
Done. Ask your first question in ChatGPT and the data comes from Shopify.
## Where this earns its keep
The stores that get the most from this are the ones where the founder still looks at orders every day. That describes most sub-$20M brands. The moment you can answer "how did yesterday go" in ten seconds instead of ten minutes, you free up an hour a day and, more importantly, you start asking questions you never asked before.
It also works well in combination with other connectors. Connect Shopify plus Klaviyo and you can ask ChatGPT about email-attributed revenue without opening either dashboard. Add Google Ads and Meta and you get blended ROAS in a single conversation. CorpusIQ's value compounds with each connector because ChatGPT can reason across them.
## What to watch out for
Two caveats worth stating plainly.
First, ChatGPT does not replace a proper data warehouse for businesses over $50M in revenue. At that scale, you need historical data, stored snapshots, and custom joins. CorpusIQ is an operational tool, not a warehouse. If you are running regression analysis on three years of order data, build a warehouse.
Second, the quality of the answer depends on how your Shopify data is structured. If your product tags are inconsistent or your collections are messy, slicing by them will produce messy results. ChatGPT is not going to fix your taxonomy. It will surface it.
## See also
- [Connect Klaviyo to Claude: Email Revenue Attribution That Actually Works](https://www.corpusiq.io/blog/connect-klaviyo-claude-email-attribution)
---
# Connect QuickBooks to Claude: Close Your Books 4x Faster
URL: https://www.corpusiq.io/blog/connect-quickbooks-claude-close-books-faster
Published: 2026-05-04
Category: connector-guide
Connector: quickbooks
Stop copying numbers between QuickBooks and spreadsheets. Connect QuickBooks to Claude with CorpusIQ and get P&L, AR aging, and cash position on demand.
import { ConnectorCallout } from "@/components/blog/ConnectorCallout";
import { FAQ } from "@/components/blog/FAQ";
Month-end close is the most avoidable time sink in a small business. A controller spends 12-18 hours copying numbers from QuickBooks into Google Sheets, chasing down expense categorizations in email, and reconciling what the P&L says against what the operator thinks happened. Most of that work is translation, not analysis.
Connecting QuickBooks to Claude through CorpusIQ cuts the translation work to zero. You ask a question in plain English, Claude pulls the data from QuickBooks in real time, and you get the answer in seconds. The numbers come from Intuit, not from a copy-paste.
## The problem with the current workflow
The standard month-end close looks something like this. Open QuickBooks, run the P&L, export to CSV, clean the CSV in Sheets, add variance columns, build a summary tab, paste into a Slack thread for the team. Run the AR aging, email the top five overdue customers. Pull cash position, compare to last month, flag anything outside the normal range. Repeat every month.
None of this is hard. All of it is slow. The finance lead becomes a data mover instead of a decision maker. When the CEO asks a follow-up question that requires slicing the data differently, the whole process starts over.
The deeper issue is that questions die before they get asked. If pulling a custom slice of the P&L takes 20 minutes, you only ask the big questions. The small questions, the ones that actually surface problems early, never get answered because the friction is too high.
## How CorpusIQ solves it
CorpusIQ exposes QuickBooks Online to Claude through the Model Context Protocol. MCP is an open standard that lets AI assistants call external tools in a structured way. When you connect QuickBooks via CorpusIQ, Claude gets a set of tools it can use to pull P&L, balance sheet, AR aging, AP aging, invoices, payments, vendors, and customer records directly from your books.
Three things make this different from generic QuickBooks reporting:
The connection is read-only. CorpusIQ uses OAuth with read-only scopes. Claude can query your books, but it cannot create, modify, or delete anything. There is no scenario where an AI typo becomes a miscategorized transaction.
Direct MCP does not retain raw customer files or full connector response payloads. When Claude asks a question, CorpusIQ retrieves the required records from Intuit and returns the answer. Operational logs may include query text, per-user tool-call metadata, and a bounded outcome summary for up to 30 days. Optional indexed search has a separate lifecycle. Claude conversation handling follows the Anthropic plan and settings you choose.
Disconnecting QuickBooks in the CorpusIQ dashboard removes the stored connection state and requires reauthorization before CorpusIQ can use QuickBooks again. Provider-side authorization remains governed by Intuit and can be managed from the Intuit admin panel.
## What you can actually do
Once QuickBooks is connected, these are real prompts you can paste into Claude. Each pulls live data from your books.
- "What is my cash position today versus 30 days ago, and what drove the difference?"
- "Show me overdue invoices sorted by age, grouped by customer."
- "Which expense categories are up more than 20 percent month over month?"
- "List every customer with an open balance over $5,000 and their payment history for the last 6 months."
- "Give me a profit and loss for Q1, then compare it to Q1 last year."
- "Which vendors have we paid the most this quarter, and which ones are on net 30 terms?"
- "Flag any transactions over $2,000 that look miscategorized based on the vendor name."
- "What is my average days-to-pay across customers, and which 5 customers are pulling the average up?"
These are not canned reports. Claude constructs the answer from the underlying data, so follow-up questions work naturally. Ask for a P&L, then ask why marketing is up, then ask which specific line items drove it. Three questions, three seconds each.
## Setup in 3 minutes
Connecting QuickBooks to Claude through CorpusIQ is a standard OAuth flow.
1. Create an account at [corpusiq.io](https://www.corpusiq.io). The Solo plan at $29.95/month includes QuickBooks plus all 40+ connectors, with a 30-day free trial.
2. From the CorpusIQ dashboard, click Connect next to QuickBooks Online. You will be redirected to Intuit to approve read-only access.
3. In Claude, add the CorpusIQ MCP server to your integrations. Full instructions are in the CorpusIQ dashboard under Setup.
That is it. Open Claude and ask your first question. No CSV export, no pivot table, no copy-paste.
## Where this pays off most
The operators who get the biggest lift from this setup are the ones with a small finance function and a lot of month-end questions. One-person finance teams, founder-operators doing their own books, and fractional CFOs supporting multiple clients all benefit because the ratio of questions to headcount is high.
It also works well for audit prep. Instead of pulling ten custom reports before a review, you connect QuickBooks to Claude and let the reviewer ask questions directly. The answers come from the source. No one is wondering whether the spreadsheet was up to date.
## What to watch out for
Two honest caveats.
First, Claude is not a replacement for a controller or a CPA. It answers questions about the data that exists in QuickBooks. If your books are miscategorized, the answers will be wrong in the same way the books are wrong. Fix the books first, then ask better questions.
Second, the tool is only as useful as the questions you ask. Teams that have never thought about AR aging by cohort will not suddenly start. The benefit shows up when you already have a reporting habit and want to stop spending an afternoon to feed it.
## See also
- [Close the Books 50% Faster: QuickBooks + Drive + Gmail Workflow with Claude](https://www.corpusiq.io/blog/close-books-faster-quickbooks-drive-gmail-claude)
---
# Five Questions to Ask Your Books Every Monday Morning
URL: https://www.corpusiq.io/blog/five-questions-ask-your-books-monday-morning
Published: 2026-04-28
Category: finance
Connector: quickbooks
A 10-minute Monday ritual that catches cash problems before they happen. Five specific finance questions, exact prompts, and what good answers look like.
# Five Questions to Ask Your Books Every Monday Morning
Most small business owners look at their books once a month, usually when something has already gone wrong. By that point, the cash problem is two weeks old and the fix is twice as expensive.
A weekly Monday ritual catches it earlier. Ten minutes, five questions. You do not need a CFO. You need the right prompts and live access to QuickBooks, Wave, or FreshBooks. Here is the ritual we recommend.
## Why Monday and not Friday
Friday feels logical. Close out the week, check the books, head into the weekend. The problem is timing. If you find a cash issue on Friday afternoon, you cannot do anything about it until Monday anyway. So you spend the weekend stewing.
Monday gives you five business days to act on what you find. Late invoice, slow customer, expense anomaly, whatever it is, you have a full week to fix it before it compounds. That is the entire game.
## Question 1. What is my real cash position right now?
Not the bank balance. Real cash position factors in pending payroll, AP coming due this week, and AR likely to land based on customer payment history.
Prompt to use:
> What is my cash position right now, accounting for AP due in the next 7 days, payroll if it falls in this period, and AR likely to come in based on past payment patterns?
The answer should give you a number you can actually rely on. Not the bank balance staring at you, which is misleading because it does not subtract obligations or add reasonable receivables. CorpusIQ runs `financial-command-center` against your books and pulls all four data points in one pass.
What good looks like. A range, not a single number. "$22,400 floor, $31,800 ceiling depending on whether the Acme invoice clears this week." That tells you how much room you have for surprises.
## Question 2. Who owes me money and how late are they?
AR aging by bucket. Current, 30 days, 60 days, 90+. Most owners think they know this. Most owners are wrong by 20 to 40 percent.
Prompt:
> Pull AR aging by bucket. Show me the top 10 overdue invoices by dollar amount, sorted by days late.
What you are looking for is the long tail. The invoice that has been sitting at 67 days that you forgot to chase. The repeat customer who is now at 45 days for the third quarter in a row, which is a pattern, not an accident. The one big invoice that is 90+ days old and is probably not going to be paid without legal action.
Tip. Run `cash-recovery-engine` immediately after this question. It drafts a 14-day follow-up sequence per overdue customer, matched to the tone of past email threads. Twenty minutes of work that recovers real money.
## Question 3. Are any expenses out of pattern this week?
Subscription creep is the silent killer. So is a vendor who quietly raised their rate. So is a duplicate charge nobody caught.
Prompt:
> Compare this week's expenses to the same week last month and the same week last year. Flag anything more than 20 percent off pattern.
CorpusIQ pulls the comparison and surfaces anomalies. Sometimes the answer is "everything is in line." Great. Move on. Sometimes the answer is "your AWS bill jumped 47 percent and your Slack subscription is now charging for users who left the company three months ago." That is $400 a month you can claw back.
This is the question that pays for the whole ritual.
## Question 4. What is my customer concentration looking like?
If your top customer represents more than 25 percent of revenue, you have a concentration problem. If it is more than 40 percent, you have a single-point-of-failure problem. Most owners do not run this number until a customer leaves.
Prompt:
> What percentage of revenue came from each of my top 10 customers in the last 90 days? Flag any customer above 20 percent concentration.
What good looks like. A ranked list with percentages and dollar amounts. The action is not always to fire the big customer. Sometimes it is to invest harder in business development to reduce dependency. Either way, you cannot manage what you do not measure.
## Question 5. What changed since last week?
The sleeper question. The one that surfaces problems you would not have thought to ask about.
Prompt:
> What changed in my finances compared to last Monday? New large expenses, new customers, lost customers, payment status changes, anything notable.
This is essentially a diff against last week's snapshot. CorpusIQ joins QuickBooks state with prior state and surfaces the deltas. New $4,200 vendor charge you did not approve. Customer who paid early for the first time in a year. Refund you forgot to issue. Whatever moved.
The answer to this question is sometimes boring. That is fine. Boring is a great answer for finance. The week you get an interesting answer is the week the ritual paid for itself.
## How to make it stick
Three tips after running this with operators for a year.
Save the prompts. Drop them in a Notion doc, a saved chat, a sticky note. Whatever works. The friction of writing prompts from scratch is what kills the habit.
Run them in the same order every Monday. Cash position, AR, expenses, concentration, weekly diff. Same order builds pattern recognition. After a month you will start to notice when something looks off before CorpusIQ even tells you.
Block 10 minutes on the calendar. Recurring weekly. Treat it like a meeting with your CFO, even if you do not have one. Especially if you do not have one.
## What to do with what you find
The ritual is useless if you do not act. Three rules.
If question 1 shows tight cash, push the AR follow-ups today. Do not wait.
If question 3 surfaces an anomaly, dig in within 24 hours. Most expense anomalies have a person or a vendor at the root, and that person or vendor will keep doing it next week if you do not address it now.
If question 4 shows rising concentration, that goes on next quarter's planning. Concentration problems take 60 to 90 days to solve.
The other questions are diagnostic. They are early warning systems. The point is to catch issues at week one instead of week six.
Ten minutes. Five questions. Every Monday morning. The cheapest finance practice you will ever build.
Internal links to add:
- "financial-command-center skill" to /skills/financial-command-center
- "cash-recovery-engine" to /skills/cash-recovery-engine
- "30-day free trial" to /pricing
External citations:
- SCORE small business cash flow research
- JPMorgan Chase Institute report on small business cash buffers
Last updated: 2026-09-03