Red Flags on Hold
Emergencies hiding in routine booking calls: found, called back by a nurse, and turned into the exam every AI booking agent must pass.
The job
Finds urgent warning signs callers mention while booking appointments, sends every missed one to a nurse for a same-day callback, and tests each AI booking agent version on real calls so the medical director can approve it before it answers patients.
The moment
Of 60,000 appointment booking calls last month, 212 callers mentioned chest pain or trouble breathing. 31 of them were simply booked in for next week.
The medical director reads one call aloud. Before the meeting ends, the head of the nurse line sends yesterday's three missed callers to a nurse for a callback.
Next, a new AI booking agent is tested on all 212 calls, played by simulated callers. It catches 209 and misses three, all in one non-English language, so it cannot be signed off.
Two weeks later the fixed version catches all 212, and the medical director approves it from a phone.
What it does
- Send to nurse callback queue
- Mark reviewed / override label
- Add call to the test suite
- Run the Go-Live Gate
- Request go-live sign-off
- Open transcript
What you see
Triage board with emergency-department colour tags per call, plus a Go-Live Gate scoreboard and a certificate ledger for AI agent versions
What it moves
The share of warning-sign booking calls escalated on the same call (target 100%), same-day nurse callbacks on missed ones, estimated misses per 10,000 booking calls, and each AI agent version's pass rate on the test calls.
Built for
- Contact-centre operations
- Compliance, complaints & legal
- QA, training & knowledge
