Why this matters
Someone asks "what did we spend on Paid Search?" and you open a spreadsheet, filter, add up a column and hope you did not miss a row. Ten minutes later you do it again for Email.
An n8n chat agent can read the sheet for you. You type the question in plain words, it reads the rows and answers. No formulas, no filters.
In this module you start from nothing. You copy n8n’s most popular chat template, use n8n’s free AI credits so you need no OpenAI account, connect your Google account, make your own copy of a sample marketing sheet, and ask it questions.
What you’ll build
A working chat in n8n that answers questions about your own Google Sheet, such as "What was the total spend for Paid Search?" followed by "Now the same for Email only".
You also learn to check what the agent read, because models can slip on sums. The last chapter shows how to swap in your own OpenAI key for when the free credits run out.
What you need
- Nothing to start the practicen8n, Google and the sheet are all practice copies. The key shown is made up.
- An n8n account, to do it for realSign up for n8n Cloud with this link (affiliate link: I may earn a commission, at no cost to you). New accounts get free AI credits, shown as "Gateway credits".
- The templaten8n template 7639, "Talk to your Google Sheets using ChatGPT". Open the page, click Use for free, then Copy template to clipboard (JSON).
- A Google account and a sheetThe lesson uses n8n’s sample marketing sheet. Open it, then File, Make a copy. Or start a blank sheet and put clear column names in row 1.
- An OpenAI API key (optional, later)Only when the free credits run out. Create it at platform.openai.com under API Keys, and add a few dollars under Settings, Billing. This is separate from a ChatGPT Plus subscription.
Concepts
3 ideas you’ll use in the practice.
A template is a workflow you paste in
The template page gives you the whole workflow as text. You copy it, open a blank workflow in your n8n, click the canvas and press Ctrl+V (Cmd+V on a Mac). All the parts appear connected.
You still have to attach your own accounts. Nodes with a warning mark are asking for a connection.
Three parts: brain, reader, memory
The OpenAI Chat Model node is the brain. The Analyze Data node reads your Google Sheet. The Memory node remembers the conversation, so "Now the same for Email only" makes sense.
Each one needs a credential, which is a saved sign-in. The model uses n8n’s Gateway credits. The sheet reader uses your Google account.
Row 1 is the labels, so keep them clear
The agent reads your column names from row 1 and the data from row 2 down. Clear headings give better answers. Click Execute step on the sheet node first: if your rows show under Output, the agent can see them too.
The agent only reads, so it never changes your sheet. It can still get a sum wrong. Open the Logs panel to see what it read, then filter the sheet yourself and compare the total.
Practice
Do every step yourself in a practice copy of the real tool. It checks your work, and “Do it for me” finishes any step you get stuck on.
Do it for real
Now repeat it in your own account. Tick each step as you go.
Quick check
3 questions to check you’ve got it.
1. You have no OpenAI account. How do you give the chat model a brain at the start?
- You cannot start without an OpenAI key
- Pick Gateway credits, the free credits n8n gives new accounts
- Use your ChatGPT Plus password
Show the answer
Pick Gateway credits, the free credits n8n gives new accounts. Gateway credits let you start with no OpenAI account. Your own key is only needed when they run out.
2. Why click Execute step on the sheet node before chatting?
- To see whether your rows and column names show up, which proves the agent can read them
- To publish the workflow
- To change the sheet
Show the answer
To see whether your rows and column names show up, which proves the agent can read them. If your rows appear under Output, with the headings from row 1, the agent can see them too.
3. The agent gives a total. What is the safe next step?
- Trust it, because it read the sheet
- Check what it read in Logs and filter the sheet to compare
- Ask the same question until it agrees
Show the answer
Check what it read in Logs and filter the sheet to compare. Models can slip on sums, so check any total before you repeat it.
IF and Switch: Route Work by Rules
Next, teach a workflow to decide: send each request where it belongs with an If node, then a Switch with named outputs.
Start module 9 →