How to Build an AI-Powered Lead Scoring Workflow in n8n

One algorithm update. One policy change. One bad quarter for a platform that isn't yours. The audience you spent years building disappears overnight.

beehiiv is what happens when you stop renting and start owning. A list that's yours. Revenue that compounds. Growth tools built in from day one.

30% off your first 3 months with code LIST30. Start building today.

Reviewing inbound leads manually is the tax you pay for not having a sales team. Every form submission looks the same when it hits your inbox — a name, a company, a vague description of what they want. You spend 45 minutes a day deciding which ones to reply to first, and you still get it wrong.

This workflow changed that. Now every inbound lead gets scored 1–10 by Gemini Flash before I see it. Hot leads land in Slack within 90 seconds. Cold leads go straight to Baserow for a weekly review pass. My actual decision-making time dropped to about 3 minutes per day.

Here's the full setup.

The node chain:

Step by step:

1. Webhook trigger
Set your Tally form's webhook to POST to this endpoint. All form fields arrive as JSON. No additional configuration needed.

2. Set node
Map the fields you care about: name, company, role, use_case, budget_range. Normalise budget to a number — Tally sends it as a string like "€5,000–10,000."

3. Company enrichment
Use an HTTP Request to Clearbit's free company enrichment API (https://company.clearbit.com/v2/companies/find?domain={{company_email_domain}}). If Clearbit returns nothing, fall through gracefully — the AI scores on what it has.

4. AI scoring
System prompt: "You are a B2B lead qualification specialist for an automation consultancy serving 50–500 employee SMBs. Score this lead 1–10 based on: company size signal, role seniority, specificity of use case, and budget alignment. Respond as JSON: {"score": 7, "reasoning": "one sentence"}."

User message: the full lead object as JSON.

5. If node
Condition: {{ $json.score >= 7 }}. True → hot path. False → nurture path.

6. Baserow + Slack (hot path)
Create the Baserow row first, then post to Slack. The Slack message format: 🔥 Score {{score}}/10 — {{name}}, {{role}} at {{company}}\n{{reasoning}}\n→ Reply within 2 hours.

7. Baserow (nurture path)
Same row creation, Status: "nurture." Review these weekly, not daily.

The whole flow runs in under 90 seconds from form submission to Slack ping.

Tally — the form builder that doesn't fight your webhook

Every other form builder makes webhooks a paid feature or buries them three menus deep. Tally gives you webhook notifications on the free plan, no questions asked.

Price: Free for unlimited forms and unlimited submissions. The paid plan ($29/month) adds file uploads over 10MB, custom domains on embed, and team collaboration — none of which you need for a lead capture form.

Setup: Create form → Integrations → Webhooks → paste your n8n webhook URL. Done in 90 seconds.

The honest tradeoff: Tally's conditional logic (show/hide fields based on answers) is functional but limited on the free plan. Don't try to build your routing logic inside Tally — that's what n8n is for. Keep the form simple and let the workflow handle everything downstream. Also: Tally's hidden fields for UTM parameter tracking break on some mobile browsers. If attribution matters to you, pass UTMs through your URL and parse them in n8n instead.

The form I use for StackSignal's services page has four fields: name, email, company, and a single open text field ("What are you trying to automate?"). Everything else the AI infers from context.

AI lead scoring is eating the SDR market — blog.close.com
Every CRM is bolting on AI scoring now. The interesting thing is not the scoring — it's what happens to the SDR role when AI pre-qualifies 80% of leads. My take: the 20% that gets escalated will be worth more, not less, and the humans handling them will need to be sharper.

n8n's AI node updates in 2025 — blog.n8n.io
The LangChain integration in n8n has matured faster than expected. Agents that would have needed 200 lines of Python 18 months ago now take 20 minutes to configure. The bottleneck is now prompt design, not plumbing.

"Qualification is a conversation, not a score" — news.ycombinator.com
A counterpoint worth reading. The argument: lead scoring optimises for fit, not timing, and most deals are lost to timing. I half-agree. Score for fit, but build a follow-up sequence for the "good lead, wrong moment" segment — don't just tag them nurture and forget them.