Quick overview
Ask your spreadsheet questions in plain language — "best month?", "revenue by country?", "rows missing an email?" — and an AI agent reads the live sheet and answers with real numbers. Conversation memory makes follow-up questions work. Five nodes, five-minute setup.
How it works
- When You Ask a Question is n8n's hosted chat — open the chat URL and type a question in plain language.
- Answer With Data, Not Guesses is an AI agent with one tool: Read the Spreadsheet. It pulls the live rows from your Google Sheet and computes the answer from real data — it never guesses from memory.
- Remember the Conversation keeps context, so follow-ups like "and the month before that?" just work.
- Answers come back in the chat with the actual numbers.
Setup
- Connect an OpenAI (or compatible) credential in OpenAI Analyst Model — any OpenAI-compatible endpoint works.
- Connect your Google Sheets account in Read Data from Sheets and pick the spreadsheet and sheet the agent should read.
- Optional but worth it: add one line describing your columns to the system message in Data-Driven Response Agent (e.g. "Columns: Date, Country, Revenue in USD") — answers get noticeably better.
- Activate the workflow and open the chat via When User Asks Question → Chat, then ask something like "what was our best month?" to test.
Requirements
- OpenAI (or compatible) API key, Google Sheets.
Customization
- Describe your columns in the agent's system prompt for noticeably better answers ("Revenue is in USD, Date is YYYY-MM-DD"). Add more sheet tools for multi-tab analysis. Add an append/update Sheets tool to let the agent write rows, not just read them. For big datasets, swap the Sheets tool for a Postgres/MySQL tool - the agent pattern is identical.
Additional info
The chat is n8n's hosted chat page. Set it to public to share the analyst with teammates who don't have n8n access.