AI Lead Qualification with Research, Scoring and Human Review
We build a workflow that researches inbound enquiries, applies your qualification rules and routes useful context to sales. Missing or uncertain information goes to a person for review.
This is one example of an AI Automation System. Scope, price, timeline and acceptance criteria are agreed before the build starts, and the qualification logic is documented so your team can adjust it.
Who this is for
This build is for B2B teams whose inbound form fills arrive with no context, so salespeople spend their mornings researching leads instead of selling and their calendars fill with calls that were never going to close.
- Marketing is generating volume but sales says the leads are unqualified.
- Reps manually Google every inbound lead before deciding whether to call.
- Hot leads sit unactioned for hours or days because nobody flags them.
- Leadership cannot tell which lead sources actually produce qualified pipeline.
What we build
- Custom ICP scoring logic: your ideal-customer criteria translated into a rubric the AI uses to grade every lead from 1 to 10.
- Automated web research: the agent collects approved public business context from the lead's website. Missing or uncertain facts remain flagged for review.
- CRM enrichment: the workflow prepares a business summary and writes agreed fields to your CRM, with source context and a review path for uncertainty.
- Priority routing: leads meeting your agreed threshold reach the assigned person through the selected CRM or notification channel.
- Lower-priority routing: enquiries outside the agreed fit criteria follow a documented path. Uncertain cases go to a person; any outbound messages need separately agreed rules.
How the workflow runs
- The researcher node: n8n opens a headless browser, visits the lead's URL and extracts key metadata and site content.
- The LLM brain: site data and form responses go to OpenAI or Anthropic with a specialised system prompt for evaluation.
- The comparison engine: the AI checks the lead against your explicit no-go list, such as students, competitors or companies under a revenue floor.
- The feedback loop: your team reviews scoring decisions in Slack. Feedback can inform prompt or rubric changes, which are tested against representative cases before release.
Inputs we need from you
- A written definition of who is and is not a qualified lead, before the build starts.
- Access to one lead source, for example your website forms.
- Access to one destination, for example HubSpot or Pipedrive, plus Slack if you want alerts.
- Your own OpenAI or Anthropic API key. Usage is billed to you and depends on volume, model, and provider.
Timeline and scope
This build is scoped and priced as an AI Automation System. Scope covers one lead source connected to one destination system, with the scoring rubric and prompts documented so your team can adjust them without us. Timeline, price and acceptance criteria are agreed in writing before the build starts.
What the AI cannot do
- It can only research leads that supply a valid business email or website URL; it cannot uncover private or hidden financial data.
- It supports qualification decisions; your team owns the criteria, exceptions and customer conversation.
- Scoring accuracy depends on the clarity of the ICP definition you provide.
Related engagements
- Use the free ICP Generator first if your ideal customer profile is not written down yet.
- Explore the automation starter examples and prices; you can book a fit call before choosing what to automate.
- Consider a Fractional CAO engagement if several workflows need owning, not just one.
Lead qualification guides
Frequently asked questions
How does the AI research leads?
The AI uses a headless browser to visit the lead's website, extracts key metadata and content from their About and Services pages, then evaluates them against your ICP scoring criteria using OpenAI or Anthropic models.
What is included in the ICP scoring?
We translate your ideal customer profile into a rubric that grades every lead from 1 to 10, covering company size, industry, services offered and any no-go criteria you specify such as excluding competitors or students.
How much do API costs run?
You supply and own your OpenAI or Anthropic API key and usage. The cost depends on volume, model, and provider.
What happens to low-scoring leads?
They follow the lower-priority route agreed before the build. Uncertain cases go to a person for review. An email nurture sequence is included only when its scope and sending rules are explicitly agreed.
Can the AI scoring improve over time?
Reviewed feedback can inform changes to the scoring prompt or rubric. We test proposed changes against representative cases before release and measure the result; improvement is not automatic or guaranteed.
How long does the build take?
It depends on the lead source and destination system. We agree the timeline, scope and acceptance criteria in writing before the build starts, once the ICP definition and system access are in place.