Hospital prospecting that actually happens — in two minutes, not an hour.

The problem

Reps were supposed to prospect hospital accounts before on-site visits — map the relevant service lines, identify the surgeons, administrators, and purchasing contacts who actually matter. In practice, the prep either took an hour of hunting across Clay, LinkedIn, and public sources, or it didn't happen at all. Most reps walked in cold.

What we built

• A prospecting agent inside Claude, wired to the rep's tools via MCP: HubSpot, Clay, the NPI provider registry, Google search, and public signal sources
• Tiered ICP discovery — rep names the hospitals, Claude pulls the HubSpot context on each account, then runs layered searches for tier-1, -2, and -3 target personas (surgeons, department heads, purchasing, etc.)
• A signal layer on top of the enrichment — buying intent, recent hires, facility news — so contacts arrive pre-qualified, not just raw names
• A human-in-the-loop approval table — rep approves contacts one by one, and the approved set flows into HubSpot with verified phone, email, and the signals that made each one relevant

How it works

A rep tells Claude "I'm visiting Hospital A, Hospital B, and Hospital C this week." Claude reads the HubSpot records for each, then searches across Clay, NPI, and public sources to build a tiered contact list per account. Each contact comes back with phone, email, role, and the signals that surfaced it. Rep approves what they want, Claude writes to HubSpot. A hospital visit goes from an hour of prep — or none — to two to five minutes.

The outcome

• 100% team adoption — all six reps use it before every on-site
• Pre-visit prep compressed from 60+ minutes (or zero) to 2–5 minutes
• 5–15 targeted contacts per hospital, live in HubSpot, before the rep walks in
• Prospecting is now a thing that happens, consistently, across the team

Stack