Automatic customer satisfaction survey with n8n and AI

Almost no unhappy customer tells you. They just don't come back, or they say it in a one-star review. And happy customers, who would be your best advertising, don't leave reviews either unless someone asks at the right moment.

In this guide you'll build an n8n workflow that sends a two-question survey the day after every job, saves the answers in Google Sheets, asks AI to summarize the comment and spot the reason, alerts you right away when someone is unhappy and invites high scorers to leave a review. You only read the alerts and call whoever needs it.

If you already reply to Google reviews with AI, this workflow works one step earlier: it catches the problem while you can still fix it privately.

What you'll build

Two small workflows that share one spreadsheet:

Sending: a Schedule Trigger every morning reads yesterday's jobs from Google Sheets and sends a Gmail email with the survey link.
Answers: an n8n Form Trigger collects the score (0 to 10) and the comment, a Basic LLM Chain summarizes and classifies it, it's saved to Sheets and an IF node decides what happens next: urgent alert, review request or nothing.

What you need

An n8n instance (cloud or self-hosted) with a public URL, so the form is reachable.
Gmail and Google Sheets connected to n8n with OAuth2.
An API key from an AI provider with a spending limit set.
A "Jobs" sheet with columns id, date, name, email, service and survey_sent.
An "Answers" sheet with id, date, score, comment, summary, reason and sentiment.
The review link for your Google Business Profile.

Step 1: the survey form

Create a new workflow with an n8n Form Trigger. Give it a friendly title ("How did we do?") and just two fields:

Score: number from 0 to 10, required. "How likely are you to recommend us?"
Comment: long text, optional. "What could we do better?"

Add a third hidden field, id, passed in the link URL (for example ?id=123) so you know which job each answer belongs to. If your n8n version doesn't support hidden fields, ask for the customer's email in a normal field and match on that.

Less is more: every extra question lowers the response rate. Two questions take fifteen seconds on a phone.

Step 2: send the survey the next day

In another workflow, a Schedule Trigger every day at 10:00. Then:

Google Sheets (Get Row(s)) on "Jobs", filtering for date equal to yesterday and empty survey_sent.
Gmail (Send) with a short, personal message:

Hi {{name}}, thanks for trusting us yesterday. Could you spare 15 seconds? It's just two questions: [survey link]. It really helps us improve.

Google Sheets (Update Row) setting survey_sent to the date, so nobody gets it twice.

Send from your normal address, not a "noreply": if the customer prefers to reply to the email, you'll see it.

Step 3: AI reads the comment

Back in the form workflow, if the comment isn't empty, pass the data to a Basic LLM Chain with a prompt like this:

You are the assistant of a small business. Read a customer's score (0–10) and comment. Reply only with JSON: {"summary": sentence of at most 15 words, "reason": one of [price, service, punctuality, quality, other, none], "sentiment": one of [positive, neutral, negative], "urgent": true if the customer mentions an unresolved problem, a serious complaint or asks for a call}. Don't invent anything that isn't in the text.

Turn on Require Specific Output Format with a Structured Output Parser so the answer always arrives with those four fields. If there's no comment, skip the AI and fill reason: none with an Edit Fields (Set) node: fewer calls, lower cost.

Step 4: save and decide

Save everything with Google Sheets (Append Row) in "Answers". Then a Switch with three outputs:

Score 0 to 6 or urgent true: instant alert on Telegram or Slack with name, service, score and summary. Call them yourself, today. A timely call turns many complaints into loyal customers.
Score 9 or 10: wait a couple of hours with a Wait node and send a Gmail invitation to leave a review with your Google link. Only ask people who already said they're happy, and don't offer anything in return: Google's policies forbid incentivized reviews.
Score 7 or 8: nothing automatic. These comments are gold for improving; read them in the weekly summary.

Step 5: weekly summary

A third mini workflow with a Schedule Trigger on Mondays: read the week's answers, compute the average score and the share of each reason, and ask AI for three sentences on what comes up most. It emails you the result. Within a month you'll know whether your issue is punctuality or price, from data instead of gut feeling.

For a fuller report, see how to build a daily AI report and switch it to weekly.

Test it safely

Add three test jobs dated yesterday with your own email.
Run the sending workflow manually and check that three emails arrive and survey_sent is filled.
Answer one survey with a 3 and "you arrived an hour late", one with a 10 and one with an 8 and no comment.
Check: urgent alert for the 3, review invitation for the 10 and just a log row for the 8.
Once it all adds up, activate the workflows.

Cost and limits

Cost: one short AI call per answer with a comment. For a small business that's a few a day; check current pricing for your AI provider and n8n.
Privacy: you store name, email and opinion. Mention it in your privacy policy, only survey real customers and respect anyone who asks not to be emailed again.
Few answers at first: it's normal for only a small share of customers to reply. A clear subject, a short message and sending it the next day help more than any trick.
AI summarizes, it doesn't decide: the urgent alert saves you reading a hundred comments, but the call to the customer is yours.

Start with ready-made templates

If you'd rather skip the basic n8n and AI setup, download the WhatsApp agent for free or check out the AgentKiosk pack: 8 ready-to-import agents, including a PRO customer support agent and a Google review reply agent. You build this survey next to them with the steps above, and it fits with both: you catch unhappy customers in time and nudge happy ones to leave a review.