Modelled on "talk to a doctor in 30 minutes," built for grid equipment: describe the question, the engine matches the best-suited specialist on duty and books inside their committed on-call window. No email archaeology, zero scheduling admin for the expert.
Loading availability…
Every expert carries a primary specialism plus secondary coverage. The engine takes the soonest free slot, then upgrades to a better-suited specialist if one is free within a day — an AIS question fetches the AIS specialist, not whoever is next.
The expert commits once to a recurring on-call window; the system books 15-minute calls inside it on a 20-minute stride — the 5-minute gap carries an AI-assembled briefing so the expert walks in prepared.
Each consultation produces two artifacts: a full case record on the linked opportunity, and a de-identified knowledge-base entry — so the next requester's triage may not need a call at all.
This is the requester flow of a validated internal prototype, running its real slot engine client-side on a seeded demo rota with fictional experts. The production architecture (Microsoft Graph free/busy, calendar-backed Teams meetings, Claude-on-Bedrock triage and briefing, Salesforce write-back) is specified in the project's planning documents mirrored under docs/ask-an-expert/.