Score inbound leads for fit before anyone calls
Each new form fill gets a fit score and a next step from a decision model, so the owner calls the strong leads first and the unclear ones are reviewed instead of being dropped.
Author's figure: Author's figure: only 47% of leads fit the client's ICP across 605M+ tokens screened (AronTomorrowLab). Reported by the builder, not measured by us.
The use case
Contact forms bring in a mix of real buyers, students, job seekers, and vendors. Jev reads each submission together with short notes about your ideal customer and answers how good a fit the lead is on a four-level scale and which next step suits it. Leads with a confidence of 0.8 or higher are tagged in the CRM with the score and suggested step. Lower-confidence leads go to a review list. Jev does not reply to anyone or remove leads. The owner or salesperson still decides who to call, what to offer, and whether a lead is worth pursuing, and Claude can draft the first reply for a person to edit and send.
Example workflow
- A new form submission arrives in HubSpot or a Google Sheet through Zapier.
- The form fields, company website text, and your ideal customer notes are sent to Jev as the state.
- Jev answers a Score question, how well does this lead fit, with the levels poor, possible, good, and strong, and a Choice question, what is the next step, with the allowed answers call this week, send information, nurture list, and not a buyer.
- Zapier writes the score, step, and confidence to the CRM record when the confidence clears the threshold.
- Low-confidence leads are added to a review view that the owner checks daily.
- For leads marked call this week, Claude drafts a short reply that the salesperson edits and sends themselves.
Tools
- Jev (TypeSafe AI)
- HubSpot
- Zapier
- Google Sheets
- Claude
- Slack
Risks and guardrails
- A good lead scored poor may be ignored — guardrail: no lead is deleted, and the not a buyer list is skimmed weekly.
- Scores can encode bias from loose ideal-customer notes — guardrail: describe fit in terms of needs and budget, never personal traits, and review the notes quarterly.
- Drafted replies could go out unchecked — guardrail: the draft lands in the CRM as a draft only, and a person sends it.
A similar pattern in a public build: https://madewithjev.com/builds/lead-outreach-scoring. See also Decision models.
Related recipe
Inbound Lead Qualification and Reply Drafts: An agent scores each inbound lead against your written ICP, checks the company's public website, and drafts a reply with a suggested next step that a rep reviews and sends; nothing goes out automatically.
Want it set up for you? Request a quote.
Sources
Written by Agents Lookup. Credits and official pages for the tools this use case names:
- Public build with a similar pattern: madewithjev.com/builds/lead-outreach-scoring
- clay-jev-people-ranker by PromptGTM on Made with Jev: madewithjev.com/builds/clay-jev-people-ranker
- clay-jev-people-ranker: original post or repo by PromptGTM: github.com/promptgtm-shared/clay-jev-people-ranker
- Only 47% of leads fit the ICP by AronTomorrowLab on Made with Jev: madewithjev.com/builds/lead-screening-605m-tokens
- Only 47% of leads fit the ICP: original post or repo by AronTomorrowLab: x.com/AronTomorrowLab/status/2102740314146341338
- HubSpot MCP server (official page): developers.hubspot.com/ai-tools/mcp (checked 2026-11-09)
- Zapier MCP (official page): zapier.com/mcp (checked 2026-11-09)
- Claude (official page): claude.ai (checked 2026-11-09)
- Slack MCP server (official page): docs.slack.dev/ai/slack-mcp-server (checked 2026-11-09)