How to qualify leads with n8n and AI (form + email → Google Sheets + alert)

If you sell services to businesses or freelancers, you know the problem: the website form and your Gmail inbox fill up with requests. Some are gold («we have three shops, £3,000 budget, want to start now»). Others are noise («we sell cheap SEO» or a bot that fills every field). If you review every one by hand, you waste hours. If you don't, the good ones slip away.

With n8n (visual automation), a bit of AI and Google Sheets you can build a filter: each contact is scored against your ideal customer, saved to a sheet, and you only get alerted on the ones worth a call.

What you'll build

For every lead, the workflow:

Collects the contact from a webhook (website form) and/or from emails arriving in Gmail (quotes, «contact us»).
Normalises the data (name, email, company, phone, budget, message) so the source doesn't matter.
The AI scores 0–100 against your ideal-customer definition and labels them hot, warm, cold or spam, with the reason and a recommended next step.
Writes a row to Google Sheets with score, category and summary.
If they clear the threshold (70 by default), it sends you an alert email with the summary and contact details.

The AI doesn't «guess» the lead: it only sees what they wrote. You set the criteria. That's what makes the workflow useful.

What you need

An n8n instance with HTTPS.
An OpenAI key (or another model compatible with the chat node).
Google Sheets OAuth2 and Gmail OAuth2 credentials.
A Google Sheet with a Leads tab and clear headers (date, source, name, email, company, phone, budget, score, category, reason, next step, summary, message).
About 40 minutes and a concrete definition of who you do and don't want to sell to.

Step 1: define your ideal customer in one place

Before you touch nodes, put this in a settings node at the start of the flow:

What you sell and to whom.
Ideal customer: sector, size, area, minimum budget, urgency.
Red flags: agencies asking to «partner», students, competitors, empty messages.
Hot threshold (e.g. 70) and the email that should receive alerts.
The Google Sheet ID (the bit of the URL between /d/ and /edit).

Be specific. «Companies that want to grow» doesn't help the AI. «Workshops or clinics with 2–10 staff in your country, budget from £500 and a clear start date» does.

Step 2: form (and optionally Gmail)

Create a Webhook node that accepts POST and copy the production URL. Your website form should send something like:

{"name":"Laura","email":"laura@company.com","company":"North Clinic","phone":"+44…","budget":"1500","message":"We want to automate bookings across 3 sites"}

Add a hidden honeypot field so bots fill it and you can discard them.

If you also want to score quote emails, connect a Gmail trigger with a tight search (e.g. subject contains «quote» or «contact», or only a specific mailbox). A code node unifies form and email into the same shape. Form only? You can drop the Gmail branch.

Step 3: the AI scores the lead

A chat node receives the lead's message plus your settings (IDEAL_CUSTOMER, red flags, what you sell). The prompt asks for a structured reply: score 0–100, category (hot / warm / cold / spam), a short reason and a next step (call today, send a proposal, archive…).

Typical weights in a prompt like this are ideal-customer fit, budget/urgency, message quality and spam signals. Tune them for your business; there is no universal formula.

Ask the AI to explain the reason. When you review the sheet, you'll see why an 82 isn't a 45, and you can tighten the prompt.

Step 4: save to Sheets and alert only on hot leads

The Google Sheets node appends a row with every field. It won't replace a CRM if you get hundreds of leads a day, but for a small business it's enough and you stay in control.

Then an IF: if the score is at or above the threshold, a Gmail node emails you the alert (name, email, phone, score, reason, next step and a short extract of the message). Warm and cold leads stay in the sheet for later review without interrupting you.

Test it with two contacts

An ideal one: several sites, clear budget, urgency («start this month»). You should get a high score, hot category, a Sheets row and an alert email.
A bad one: generic pitch, no budget, or obvious spam. It should land as cold or spam, with a row but no alert.
Check that source (form / email) and the other columns fill correctly.

Honest limitations

The score is a guide. The AI only sees what the lead wrote. A strong customer who writes poorly can score low; someone who pastes a «perfect» message can score high. Review the first 20–30 leads and refine IDEAL_CUSTOMER.
The Gmail trigger usually polls about once a minute. The same email isn't processed twice, but changing the search can pick up recent messages from the last few days.
Very long messages are worth truncating (to a few thousand characters) so you don't burn tokens or flood the prompt.
It doesn't replace commercial judgement. It's a filter and a filing cabinet, not an autonomous salesperson.
The webhook is public. The honeypot helps; if spam still arrives, add header auth or a captcha on the form.

Ideas to go further

Auto-reply to hot leads («Thanks, we'll call you today») with a Gmail node after the IF.
Replace or duplicate Sheets with HubSpot, Pipedrive or Notion (n8n has nodes for them).
Link the form from your WhatsApp AI agent so people who ask about pricing get the form and arrive already scorable.
Pair it with Google Calendar bookings when the next step is a demo or first meeting.

Shortcut: the agent already built

This workflow is one of the eight agents in the AgentKiosk pack: an n8n template ready to import, settings in one node, scoring prompt, sheet headers, alert email and a PDF guide. It has been imported and run end-to-end in n8n with sample data and simulated services; run your own tests with your sheet and Gmail before you use it with real customers.

To start for free, download the WhatsApp agent and also read 7 AI automations for small businesses.