AgeBridge Guided ProjectReviewed by AgeBridge
AI Lead Qualification Assistant
An assistant that reads every inbound lead from both channels, scores it against a written ideal customer profile, explains the score in one sentence a human can argue with, creates the CRM record, and drafts the first reply.
ML
Maya Levin
AI automation builder
Business problem
A 9-person B2B team received around 120 inbound leads a month through a website form and a shared inbox. Roughly two thirds were students, job seekers or out-of-scope requests, and the ones worth talking to waited an average of 31 hours for a first reply because someone had to read everything manually.
Solution
An assistant that reads every inbound lead from both channels, scores it against a written ideal customer profile, explains the score in one sentence a human can argue with, creates the CRM record, and drafts the first reply.
What was built
- •One intake that merges website form submissions and the shared inbox
- •An ideal customer profile written as ten checkable rules, not a vibe
- •Scoring step that returns a 0–100 fit score plus a one-line reason
- •CRM record creation with score, reason and source attached
- •Slack alert for anything above 70, draft reply for everything else
Architecture / workflow
- 1Lead arrives from the website form or the shared inbox
- 2Company domain is enriched: size, industry, country
- 3Model scores fit against the ICP rules and writes the reasoning
- 4Record created or updated in HubSpot, deduplicated by email domain
- 5Hot leads alert Slack; the rest get a drafted reply awaiting approval
Integrations
HubSpotSlackGmail
Tools / stack
OpenAIn8nPostgres
Screenshots
Screenshot slot
Screenshot slot
Demo
60 second walkthrough: one lead from form to Slack alert
Results
Reported by the builder: first reply time dropped from about 31 hours to under 10 minutes.
Evidence
- Demo video · One lead, end to end (0:58)60 seconds showing the thing actually running. The single strongest proof.
- Workflow export · n8n workflow exportThe automation or agent graph, exported.
- Architecture notes · Scoring rules and data flowHow data flows and why you designed it that way.
- Screenshots · Slack alert and CRM recordReal screens from the build, not mockups.
Proof
Proof score
70/100Add live link — A working link a buyer can click and try.
An AgeBridge reviewer checked the evidence behind this project.
Decisions and tradeoffs
- •Scores are never hidden. Every score ships with a reason so the sales team can correct the rules instead of distrusting the tool.
- •No auto-send. The assistant drafts; a human presses send. That kept the team willing to use it.
- •Deduplication happens on company domain, not email, because three people from the same company often fill the form in one week.
What went wrong
The first version scored on job title alone and rated every junior employee at a target company as a bad lead. The rules were rewritten to score the company first, the person second.
What I'd do next
Feed back which leads actually closed, so the rules can be tuned against outcomes instead of opinions.
Need something like this?
Talk to the builder behind the project, or build your own version.