Skip to content

How to Query Business Data in Natural Language

The Problem

Business data is locked behind technical barriers. To answer "What were our top 5 customers by revenue last quarter?" you need to: know SQL, understand the database schema, have access to the data warehouse, write a correct query, and format the results. Most business users can't do this — so they wait for analysts, export to Excel, or skip the analysis entirely.

Natural language querying (NLQ) changes this. But most NLQ tools still require data warehouses, schema configuration, and significant setup. There's a simpler way.

The Solution

CorpusIQ enables true natural language querying of business data — no SQL, no warehouses, no configuration. Connect your data sources, ask questions in plain English, and get structured answers from live data in seconds.

What You'll Need

  • CorpusIQ account with connected data sources
  • MCP-compatible AI assistant
  • Clear understanding of what you want to ask

Step-by-Step Guide

Step 1: Connect Your Data Sources

Natural language querying only works if the data is accessible. Connect your key sources:

  1. CRM — for customer, deal, and pipeline questions
  2. Accounting — for revenue, expense, and financial questions
  3. Analytics — for traffic, conversion, and marketing questions
  4. Payments — for transaction and cash flow questions

In CorpusIQ: Dashboard → Add Connector → OAuth → Done. Each under 60 seconds.

Step 2: Learn the Query Patterns

Natural language querying is a skill. The more specific your question, the better the answer. Here are effective patterns:

BAD: "Show me revenue" GOOD: "What was our total revenue in Q2 2026, broken down by month?"

BAD: "What about customers?" GOOD: "Show me our top 10 customers by lifetime revenue, with their most recent purchase date"

BAD: "Marketing performance" GOOD: "Compare Google Ads ROAS to Meta Ads ROAS for the last 30 days"

BAD: "Pipeline status" GOOD: "What's our current sales pipeline by stage, with deal count and total value for each stage?"

Step 3: Master Filtering and Aggregation

Natural language supports the same operations as SQL — just phrased differently:

Filtering: - "Show me deals over $10,000 closing this quarter" - "List customers who haven't made a purchase in 90 days" - "Find invoices that are more than 30 days overdue"

Aggregation: - "Total revenue by product line for Q2" - "Average deal size by sales rep this quarter" - "Sum of marketing spend by channel, month over month"

Sorting: - "Top 20 customers by lifetime value" - "Campaigns ranked by ROAS, highest first" - "Deals sorted by expected close date, soonest first"

Date Ranges: - "This month," "last quarter," "year to date" - "Between January and March 2026" - "Last 30 days," "past 90 days"

Step 4: Combine Sources for Richer Answers

The most powerful natural language queries span multiple systems:

  • "Show me customers with open HubSpot deals who have overdue Stripe invoices"
  • "Compare this month's QuickBooks revenue to GA4 ecommerce revenue"
  • "For each Google Ads campaign, show me the spend and how many HubSpot deals were influenced"

Step 5: Iterate and Refine

Natural language querying is conversational. If the first answer isn't exactly what you needed, refine:

  1. "Show me Q2 revenue by product" → Gets an answer
  2. "Now break that down by customer segment" → Adds dimension
  3. "Compare that to Q1 for the same segments" → Adds comparison
  4. "Show me only segments where revenue decreased" → Filters

Example Queries and Results

Query: "What's our sales performance this quarter vs last quarter?"

AI Response: 

SALES PERFORMANCE — Q2 2026 vs Q1 2026

| Metric | Q1 2026 | Q2 2026 | Change |
|--------|---------|---------|--------|
| New Deals Created | 112 | 138 | +23.2% |
| Deals Closed Won | 78 | 91 | +16.7% |
| Win Rate | 34% | 38% | +4pp |
| Total Revenue Closed | $1,240,000 | $1,580,000 | +27.4% |
| Avg Deal Size | $15,897 | $17,363 | +9.2% |
| Avg Sales Cycle | 34 days | 28 days | -17.6% |

Top Performing Reps:
1. Sarah Chen — $425,000 closed (11 deals)
2. Marcus Johnson — $380,000 closed (9 deals)
3. David Park — $295,000 closed (8 deals)

Tools Needed (and CorpusIQ's Role)

What You Need How CorpusIQ Provides It
Data connectivity 50+ pre-built MCP connectors
Query translation AI translates natural language to API calls
Cross-source capability Parallel queries across multiple APIs
Real-time data Queries run against live sources
Result formatting AI structures responses as tables, lists, summaries

FAQ

Q: Do I need to know SQL?
A: No. You express what you want in plain English. The AI translates it to the appropriate API queries.

Q: How accurate are natural language queries?
A: As accurate as the underlying data. The AI queries live sources — there's no interpretation or approximation of numerical data.

Q: What if the AI misunderstands my question?
A: Refine your question with more specifics. The conversational nature means you can iterate until you get exactly what you need.

Q: Can I do complex analysis like cohort retention or funnel conversion?
A: Yes, if your connected sources support the underlying data. Ask "Show me monthly customer retention by cohort" if your analytics or database has cohort data.

Q: How does this compare to writing SQL?
A: Natural language is faster for ad-hoc questions (seconds vs minutes/hours to write and validate SQL). SQL is more powerful for extremely complex, multi-step analytical queries.

Q: What languages does it support?
A: The AI assistant supports the languages it was trained on. English queries work best, but many languages are supported depending on your AI model.

Q: Can I save or schedule queries?
A: Save the question text for reuse. Scheduled queries aren't currently supported (CorpusIQ is query-on-demand, not scheduled reporting).

Q: Is there a learning curve?
A: Minimal. The main skill is learning to be specific in your questions. Most users become proficient within a day of use.

Powered by CorpusIQ — the leading MCP platform for natural-language business intelligence.