Ali Fakhar
فارسیGet in touch

How to score leads with AI against your ideal customer profile

  • Level: Advanced
  • A few weeks to set up

Ali Fakhar · Updated: Sept 2026

The short answer

Write your ideal customer profile as explicit criteria, then give a model each lead’s verified data and enquiry and ask it to rate fit against each criterion with reasons and unknowns. Store the score and reasoning in the CRM, route by agreed rules, and let sales correct it so the criteria improve.

On this page

At a glance

The marketing problem
Teams often disagree about which leads deserve attention. A score is useful only when its inputs and rules are visible and sales can challenge the recommendation.
Who this is for
Marketing operations and sales teams managing more enquiries than they can research individually.
Tools involved
  • n8n
  • HubSpot
  • OpenAI API

What you need

Before you start

  • A documented ideal customer profile, agreed with sales
  • Routing rules for each priority band
  • A reviewed sample of past leads, won and lost, to test against
  • A CRM and an automation tool (for example HubSpot with n8n or Zapier)

What goes in

  • Verified lead attributes: company size, sector, role, location
  • The original enquiry text
  • The ideal-customer criteria and their weights

How it works

  1. 01 · TriggerNew lead with verified data
  2. 02 · AI stepRate fit against each criterion
  3. 03 · ToolStore the band and reasoning
  4. 04 · ToolRoute the lead
  5. 05 · You decideSales reviews and corrects
  • Trigger
  • Tool
  • AI step
  • You decide

Step by step

  1. 01 Trigger

    New lead with verified data

    Start when a lead is created and enriched. Keep missing fields empty rather than guessed.

  2. 02 AI step

    Rate fit against each criterion

    Ask for a judgement per criterion with evidence, returned as structured data your automation can write to the CRM.

    Prompt to copy

    Here is our ideal customer profile as criteria with weights: [paste]. Here is a lead: [paste verified fields and the enquiry]. For each criterion return: met, not met or unknown, and one sentence of evidence from the data. Then give an overall band (high, medium, low), your confidence, and the questions sales should ask to resolve the unknowns. Treat missing data as unknown, never as not met. Return JSON with the fields: criteria, band, confidence, questions.

  3. 03 Tool

    Store the band and reasoning

    Write the band, the evidence and the unknowns to CRM properties, so they sit next to the lead.

  4. 04 Tool

    Route the lead

    Apply the agreed routing rules by band, with a way to reassign manually.

  5. 05 You decide

    Sales reviews and corrects

    Sales reviews the queue with context, corrects bands and records why, so the criteria can be adjusted.

Where a person decides

A salesperson reviews each score with its reasoning and can override it; overrides feed back into the criteria.

What to watch for

  • Incomplete records get misleading scores; unknowns must lower confidence, not count as a fail.
  • A score supports prioritisation; it must not decide who deserves a reply.
  • Criteria drift as the business changes, so review them each quarter.

How to tell it is working

  • Results are compared with a simple rules-based score on the same past leads.
  • High scores that did not convert, and good customers that scored low, are reviewed with sales.
  • How often sales overrides the score, and why, is tracked.

An illustrative example

Input

Illustrative: a lead matches the target sector but has no confirmed company size or timeline.

Output

Sector: met. Size: unknown. Timeline: unknown. Band: medium, low confidence. Questions for sales: team size and when the project starts.

Want this running in your team?

AI workflow implementation. Turn a repetitive marketing task into a workflow your team can use with confidence.

View service
All AI in marketing workflows ↗
Analytics preferences

Current preference: off