Automated ICP scoring engine at webAI

Inside the first six months at a VC-backed AI startup, stood up a closed-loop ICP scoring engine that enriches and routes every inbound lead automatically.

Context

webAI's inbound was arriving in HubSpot with almost no context. Sales was burning cycles on manual firmographic research and ICP judgment calls, and speed-to-first-touch was slipping as inbound volume grew.

There was no standardized ICP rubric, which meant routing depended on tribal knowledge and lead quality couldn't be compared across reps or campaigns.

Approach

1. 1

   Capture

   New inbound lead lands in HubSpot CRM — whether via form, meeting booking, or API.

2. 2

   Discover emails

   Lead is pushed into Clay, which runs a layered waterfall of email-discovery providers to surface verified contacts on the lead's organization.

3. 3

   Add firmographic context

   Apify runs inside Clay to pull firmographic data — industry, size, intent signals, and related context — on top of the contact layer.

4. 4

   AI cleanup with Claude

   Claude runs inside Clay to clean up the enriched record: normalize company names, deduplicate against existing HubSpot data, and reliably identify the headquarters location.

5. 5

   Sync back

   The fully enriched record is written back to HubSpot, mapped to typed properties so downstream workflows can act on it.

6. 6

   Score & route

   HubSpot workflow rules evaluate the enriched lead against the ICP rubric, assign an ICP tier, and auto-route in-fit leads to the right AE. Out-of-ICP leads route to nurture.

Result

Consistent ICP application across every inbound lead, measurable speed-to-first-touch improvement, and a sales team that stopped doing the research work.

The same enrichment layer became the source-of-truth for downstream outbound, reporting, and forecasting — one model, fed once, reused everywhere.