Claude reads the full context behind every lead, then tells you whether to call.

Claude doesn't just look at the form fields. It reads the contact's recent LinkedIn posts, the company's last 90 days of press, hiring signals, your prior touches with anyone at the account, and the prose of your closed-won notes. Long context means it reads the whole stack at once — no lossy summarization step in the middle.

A rules engine grades fields. A vector model finds 'similar' deals by embedding. Claude reads the actual prose — your win-loss notes, the rep's call summary, the prospect's own words — and reasons about whether the patterns match. It's slower and more expensive than either, and worth it for deal sizes that justify a sales motion.

It says so. The score comes with a confidence band, the reasoning, and the specific signal that's missing. A '60, low confidence — title is ambiguous, verify on the call' is more useful than a confident-looking 85 that hides the same gap.

HubSpot, Salesforce, Pipedrive, Attio, and Close. Enrichment from Clay, Apollo, ZoomInfo, or LinkedIn Sales Navigator. Custom CRMs work via REST or the 1,500+ app library.

Point Claude at the last 6–12 months of closed-won and closed-lost. It reads them in full, surfaces the patterns it sees, and asks you to confirm the ones that match your judgment. The rubric is editable in plain English at any time.