Inbound lead → qualified opportunity
A new lead lands in the CRM, an AI agent qualifies against the ICP, and the right human gets a Slack or email ping with full context.
What wasn't working
Sales reps triaged every inbound form by hand, losing hours per week to unqualified traffic.
Response SLA slipped past 4 hours, and hot leads were going cold before anyone opened them.
Qualification notes were inconsistent — every rep captured different fields, breaking downstream reporting.
What we needed to be true
Respond to every inbound lead in under 2 minutes.
Score against a shared ICP so reps only see qualified opportunities.
Push structured notes into GoHighLevel and Slack automatically.
How it was built
Capture & enrichment
A single webhook endpoint receives the form payload, enriches company data via Clearbit, and normalizes it into a canonical schema.
AI qualification agent
GPT-4o scores the lead against a JSON-schema ICP prompt, produces a short rationale, and tags the record with confidence + intent.
Handoff & notification
Qualified leads land in GoHighLevel with the full context; the owning rep gets a Slack DM with the summary and a one-click booking link.
How the pieces connect
Implementation timeline
- Phase 01
Discovery
Mapped the current triage workflow with the sales team, captured ICP rules.
- Phase 02
Prototype
Shipped a working n8n prototype in 5 days for internal testing.
- Phase 03
Hardening
Added retries, dead-letter queue, and Sentry alerts for failed enrichments.
- Phase 04
Rollout
Migrated production traffic behind a feature flag with a 48-hour observation window.
Stack
What ships to the user
Sub-2-minute response
Median lead-to-Slack time is 47 seconds.
Structured scoring
Every lead lands with a confidence score, rationale, and next-action.
Human-in-the-loop
Low-confidence leads route to a review queue instead of auto-replying.
Before → after
Manual triage taking 4+ hours to respond
Automated triage responding in under 2 minutes
Free-text notes, no reporting
Structured JSON, dashboards on day one
Business impact
Let's map your workflow and see what's worth automating first.
Lessons learned
Prompt-only scoring is fragile. Wrapping the LLM in a JSON schema and validating output catches 90% of edge cases.
A cheap enrichment call before the LLM saves 30% on token spend and improves score quality.
Future improvements
Add auto-drafted reply for high-confidence leads (human-approved before send).
Route by territory + product line using the same scoring rationale.
Common questions
More work like this
Research → draft → review → publish
4× editorial output with the same headcount.
Portal scrape → clean → load
Removed 20+ hrs/week of manual portal work.
Have a workflow that looks like this?
Bring me one real bottleneck. I'll tell you whether it's worth automating, and what it would take.