When a board is asked to approve AI in repairs, complaints or customer service, the paper often describes the tool. The more useful conversation is about the work the tool will touch, and who remains answerable when it is wrong.
On 24 November I will be at Homes UK at ExCeL London, chairing a panel on how operational housing teams are using AI in practice. The questions below are the ones I would want a board or executive team to have asked before that kind of work goes live.
The context is already demanding
Social landlords in England are already working to fixed legal timeframes. Awaab's Law has been in force since 27 October 2025 for emergency hazards and for significant damp and mould hazards. Landlords must investigate a potential emergency hazard and, where required, carry out the safety work within 24 hours of becoming aware of it. Potential significant hazards must be investigated within 10 working days, with a written summary for the tenant within 3 working days of the investigation concluding, and relevant safety work within 5 working days of that conclusion.
From 30 November 2026, those duties extend to further hazards, including excess cold and heat, falls, structural collapse, fire and electrical risks, and domestic hygiene. That is set out in the government's Phase 2 guidance.
Separately, the Housing Ombudsman's Complaint Handling Code has been a legal requirement for member landlords since 1 April 2024, under the Social Housing (Regulation) Act 2023. Boards already appoint a Member Responsible for Complaints, and they already have to show, each year, that the Code is being followed.
I mention this because AI in frontline housing will sit inside those duties, whether the business case says so or not. A tool that helps staff meet a timeframe is useful. A tool that buries a report, mis-labels a complaint or leaves no record of why a case was prioritised creates risk for the tenant and for the board.
Questions worth asking before you approve it
What decision is the system allowed to make? Be precise. Drafting a summary of a repair report is one thing. Setting the priority of that report, or sending the tenant a message that the case is closed, is another. Write down what the system may do and what a person must still do.
If this is wrong, who answers? One named executive, not "the system" and not "IT". That person should be able to explain a case to the regulator, the Ombudsman and the tenant. If they cannot, the design is not ready.
Can we reconstruct what happened? For any case the tool has touched, can a competent colleague see what was reported, what the system suggested, what the member of staff decided and when? If you cannot reconstruct it, you will struggle to defend it.
"If they cannot explain a case to the regulator, the Ombudsman and the tenant, the design is not ready."
Does this help staff do the work they are already accountable for? The test is time, quality and whether the people who run repairs and complaints would use it on a busy Tuesday. If it adds a step they will work around, it will not last.
How will tenants know a person is still accountable? If a resident reports damp, they need to know a person has read it, a person has judged the risk, and a person will come back to them. If the first response feels automated and the follow-up never comes, trust falls quickly, and so does confidence in the wider service.
Have the people who do the work helped design it? A vendor demonstration is not evidence. Ask to see how the tool behaves with your own messy cases, including the ones that do not fit a template, and ask the frontline team what they would refuse to let it do.
Build this into the approval
A board does not need to become a technology committee. It needs to be satisfied that AI in frontline services has an owner, a limit, a record and a way to stop it. Those are the same disciplines you already apply to safety, complaints and financial authority. They belong in the approval. Waiting until after the first incident is too late.
Free guide
The AI Leadership Briefing
Ten questions every board should ask about AI before the next budget round, with what a good answer sounds like and the warning signs to look for.
AI Board Pack
Take the ten questions into your next board meeting
- AI budget paper template (Word)
- AI readiness scorecard (Excel)
- 60-minute board session deck (PowerPoint)
- Board AI checklist (Word and PDF)
£9 + VAT
Instant download
By buying you agree to the terms of sale.