Why Safety Latches Matter in AI Receptionists

When an AI receptionist takes a maintenance call, it's not just recording words — it's making decisions about what happens next. For UK property teams, those decisions can involve gas safety, electrical hazards, safeguarding concerns, and high-value repairs that require authorised human clearance.

The Problem with Pure AI Triage

Most AI receptionists on the market today operate as black boxes: they transcribe, categorise, and route automatically. But maintenance intake isn't a simple classification task. A tenant reporting "smell of gas" isn't the same as "dripping tap" — and treating them identically creates liability, safety risk, and regulatory exposure.

What Safety Latches Do

Safety latches are fixed rules that hold specific categories of work for human review:

• Gas and electrical work: Any report mentioning gas smell, electrical sparking, or carbon monoxide holds for Gas Safe / NICEIC registered engineer clearance

• Safeguarding: Reports involving vulnerable adults or children escalate to designated safeguarding leads

• High-value work: Repairs above a configurable threshold (default £500) require authorised sign-off

• Structural concerns: Damp, subsidence, roof integrity — held for surveyor review

The AI extracts structured facts. The safety latch decides the workflow. The authorised person makes the decision. Every step is auditable.

Not a Live Phone Service

It's important to be clear: MaintenanceOS is pre-launch. The AI receptionist capability is under testing. We do not offer a public always-on phone service. The safety latch architecture exists because we believe AI should assist — not replace — human judgement on safety-critical decisions.