Ask a head of repairs what their organisation is doing about AI and the answer often comes back a little weary. There's a working group. There's a strategy document somewhere, and possibly a pilot running over in customer services. What there usually isn't is a change that anyone on the repairs team would notice on a Tuesday morning, which is where most of the scepticism comes from.

Across UK property management, around three quarters of firms say they use AI in some form, but only about 1% describe it as fully integrated into how they work, and only 7% rank technology upgrades among their priorities for 2026. Social housing sits in much the same place, and the cost of the lag is easier to see there because the operational pressure is already so visible. Repairs and maintenance spend reached a record £10bn in 2024-25, up 13% on the previous year, with providers forecasting £10.7bn over the next twelve months in the Regulator of Social Housing's quarterly survey. Over the same period the Housing Ombudsman investigated 43% more repairs complaints and upheld 73% of the ones it decided.

The teams getting something useful out of AI have generally picked one job and done it properly, rather than waiting for an organisation-wide answer. Job intake is usually the honest place to start. Most repairs decisions are made on partial information gathered in a three minute phone call, by an adviser who has no way of knowing whether the resident has a health condition that changes the priority, and no photo of what the resident is trying to describe. Improve the quality of what arrives at the front door and the effect carries all the way through: what gets prioritised, who gets sent, what they bring with them, and whether the job is finished on the first visit. That is a narrow enough change to test on one patch and measure honestly within a quarter.

Alix works on exactly that point. It structures the information coming in, flags resident vulnerabilities so they travel with the job instead of sitting in a separate record, and lets residents send photos or short video before anyone commits an operative to a visit. Repairs leaders get a triage view of the whole workload rather than only the jobs shouting loudest, and every photo, note and decision stays in one place, which matters a great deal when the Ombudsman asks what happened and when. None of that needs a five year plan behind it. It needs one process worth improving.

AI in repairs doesn't have to feel abstract. The useful version tends to be fairly unglamorous: better information, earlier, in the hands of people who are already doing the work. If your working group is still looking for somewhere to start, the front door is usually it.