AI Lead Scoring Workflow
An AI-assisted workflow that scores and routes inbound leads in real time, reducing manual qualification time by 80% and surfacing the best opportunities before they go cold.
- sales
- AI
- CRM
- lead scoring
The problem
The sales team was spending 3–4 hours per day manually reviewing inbound form submissions to decide which ones were worth a same-day response. Volume had grown to ~80 leads per week. Most were noise. The signal was getting buried.
The delay between submission and first contact was averaging 22 hours. For the deals that eventually closed, that delay consistently showed up as a friction point in retrospective calls.
The solution
An automated lead qualification and routing workflow that:
- Enriches every inbound lead with company and contact data
- Scores the lead against a weighted rubric using Claude
- Routes high-confidence leads to the right rep via Slack — within 5 minutes
- Logs the full scoring rationale to HubSpot for visibility
Workflow architecture
Scoring rubric
Claude receives the enriched lead profile and evaluates it against a structured prompt covering:
| Dimension | Weight | What is evaluated |
|---|---|---|
| Company fit | 30% | Industry, size, growth signals |
| Role fit | 25% | Seniority, decision-making authority |
| Intent signals | 25% | Form answers, page history |
| Timing signals | 20% | Recent funding, hiring patterns |
The output is a score from 0–100 with a written rationale. The rationale is what makes the system trustworthy — reps can see why a lead was flagged, not just that it was.
Slack notification format
High-scoring leads trigger a Slack message in the relevant rep’s DM:
🟢 New lead — Score: 87/100
Company: Meridian Analytics (Series B, 120 employees)
Contact: Sarah Chen, VP of Operations
Source: Demo request form
Why this scored high:
• Operations leader at a funded SaaS company in the target vertical
• Mentioned "evaluating alternatives to current stack" in form
• Company hiring 3 operations roles — growth signal
Next step: Respond within 2 hours.
[Open in HubSpot →]
Results
After 90 days in production:
- Manual qualification time reduced from 3–4 hrs/day to ~30 min/day
- Average time-to-first-contact for scored leads: 8 minutes (down from 22 hours)
- Rep feedback: scoring rationale is trusted and used in first conversations
- False positive rate (high score, wrong-fit): ~9%
The 9% false positive rate is a known tradeoff. The current rubric is optimized for recall over precision — we would rather route a few marginal leads than miss a good one. This is a configurable dial.
What we would change
The enrichment step adds ~4 seconds of latency and fails on roughly 12% of submissions (typically small companies or personal emails). When enrichment fails, the workflow proceeds with raw form data. Scores are slightly less reliable in this path.
A future iteration would add a secondary enrichment source as fallback, and potentially a lower routing threshold for unenriched leads (flag as “unverified” in Slack).