Features

A staffing engine – not a chatbot

Everything needed to turn an open shift into a filled one, behind REST + Webhooks. Industry-neutral: a shift is just time + role + candidate list.

See the API reference

Strict call logic

The first version of Tidvis Staff follows a strict script. The AI decides nothing on its own – it just classifies the answer and lets the coordinator make the call.

1. Call candidate.
2. Present the shift briefly.
3. Ask if the person can work.
4. Allow only clear outcomes: yes, no, maybe, question, unclear.
5. On yes: pause, send to coordinator, or auto-book if rules allow.
6. On no or no answer: call the next candidate.
7. On unclear: mark requires_review.

AI output – structured, no free-form chatter

The classification is always returned in the same shape, regardless of how the answer was phrased.

{
  "intent": "accepted",
  "confidence": 0.94,
  "needsHumanReview": false,
  "candidateQuestion": null,
  "normalizedAnswer": "Can work the full shift"
}

acceptedSays yesdeclinedSays nomaybeUnsurewrong_personWrong personasked_questionAsks a questionneeds_callbackNeeds to call backunclearUnclearangry_or_sensitiveUpset / sensitive

Industry-neutral contract

Tidvis Staff doesn't care what industry you're in. A shift is time + role + location, and a candidate is name + phone + priority. The same engine fits personal assistance, home care, restaurants, warehouses, events, cleaning, security and transport.

· shift = { id, startTime, endTime, role, location? }
· candidate = { id, name, phone, priority }
· No baked-in assumptions about clients, guests, patients or industry
· Customers can choose their own call script per request

Key safeguards – built in from day one

· Calls are recorded only when you have a legal basis and clear notice.
· Transcripts can be disabled per customer.
· The AI must not disclose sensitive details about clients/guests/patients beyond what's necessary.
· Candidates can say 'don't call again' – the decision is logged and respected.
· Every decision is captured in an audit trail.
· Auto-booking is optional and off by default.
· On low confidence, the coordinator reviews before anything is booked.