Repairs reporting that keeps dangerous cases in front of a person

• The first report isn't recognised for what it is. A resident describes mould in a child's bedroom, or a smell near the boiler, in everyday words. If the person taking the call doesn't hear the risk, the clock starts without anyone noticing.
• Out-of-hours cover is thin. Evenings and weekends rely on a call centre or an on-call rota with less context and less time.
• Promises outrun reality. "Someone will be with you tomorrow" turns into a complaint when nobody comes.
• The record is incomplete. Notes are brief, times are missing, and what the resident was advised isn't written down.

It captures the report properly. The assistant collects the caller's details and turns what they say into a structured repair request, so the case starts with the facts.

Fixed rules decide, not the AI. Safety, the progress of the case and approvals are decided by fixed rules on our server. The AI can't override them, however the conversation goes.

Safety-critical reports are held for a person. Reports our rules flag as safety-critical can enter escalation and human review. They're latched, and only an authorised person can clear them. Our safety rules page explains how this works, and why the list of report types isn't published yet.

999, never 911. Where there's an immediate danger to life, the assistant's guidance is 999, never 911.

An audit trail on the case. Consequential actions leave audit evidence, and the system only reports something as done when there's evidence it succeeded.

The assistant must not invent a booking or appointment time, a contractor dispatch, a refund, compensation, a waived fee, account information or a safety reassurance. If nothing has been arranged, it won't say it has. Not a live always-on phone.

It has no authority to move money (no client-money authority). Quotes and invoices go through checks, but there's no payment, refund, compensation or fee-waiver execution in the system. A resident can ask to be transferred to a person (live transfer handling is still under testing), and a technical failure is built to fail safe rather than pass the call on unplanned.

We'll be straight with you: no software can meet a legal duty for you, and we won't claim MaintenanceOS makes you compliant. The duties, and the judgement calls, stay with your organisation.

What the Awaab's Law timescales depend on, in practice, is knowing when a report arrived and what it said. MaintenanceOS is designed to capture reports in a structured way, keep safety-critical ones in front of an authorised person, and leave a record of what happened. Our guides set out the social housing timescales with official sources, including the second phase due on 30 November 2026, subject to Parliament's approval.

MaintenanceOS is built for many organisations on one system. Your properties, residents, contractors, work orders, jobs, communications, exports, action evidence and call records sit in your organisation's own space. Isolation controls are in the code and covered by tests. Requests from the phone provider are signature-checked.

We know housing providers ask about certifications, penetration testing and single sign-on. We don't hold certifications such as ISO 27001 or SOC 2, and items such as single sign-on, multi-factor authentication, penetration testing, monitoring and legal, privacy and telecoms review are still planned before production. We'd rather tell you that now than in procurement.

At our 30 September 2026 product checkpoint, all 196 targeted claims-and-safety tests passed, covering explicit transfer, the "never invent" rules, organisation isolation, case lifecycle, decision rules and invoice rules. Local acceptance, including a full test pass and backup and restore, was evidenced on 10 September 2026. These are targeted, automated tests run against the product code in a controlled test set-up, not on a live service. They show specific rules behaved as designed in testing, not that every rule will hold in every real situation, and voice intake is still under testing.

We're pre-launch. Voice intake is still under testing, and we're taking early-access requests from housing providers. In a demo we'll take a sample report through intake, escalation and clearance, using sample data.