Residents are using AI to write complaints to their council.
Councils are using AI to read them.
Somewhere in the middle, a member of staff is trying to work out what actually went wrong.
It’s already in Wiltshire Council’s latest annual complaints report.
Wiltshire received 783 formal complaints in 2025-26. That’s up 27% on the year before, and more than double the total four years ago. Its complaints team also handled 2,384 service requests, a 40% rise.
And it isn’t just Wiltshire. Complaints received by the Local Government and Social Care Ombudsman rose 33% nationally. Ashford Borough Council’s 2025-26 complaints report to Cabinet also addresses AI-generated complaints.
The council’s report notes something every complaints manager in the country will recognise. More residents are using AI to draft their complaints. Those complaints are often much longer. Some contain factual errors, misunderstandings about which body is responsible for what, or legal references that don’t apply.
That makes it harder to find the core issue.
So Wiltshire is doing the sensible thing. It is using AI to help with the administration, so officers can spend their time on investigation and resolution. It will develop guidance on responsible use during 2026-27, with human judgement kept at the centre.
That’s a good, honest, well-governed response. And publishing the numbers openly is exactly what the sector needs more of.
But it raises a question every council should be asking this autumn.
Where is your AI pointed?
Because there’s a version of the next three years that looks like this:
Residents’ AI writes longer complaints. The council’s AI summarises them. Residents’ AI drafts the escalation. The council’s AI drafts the reply.
Both sides get more productive. Nobody’s bin gets collected any sooner.
That’s not transformation. It’s an arms race at the counter.
A complaint is one of the most expensive forms of what systems thinkers call failure demand (a term coined by John Seddon), contact caused by something not being done, or not being done right, the first time. By the time it reaches a complaints team, the resident has usually already phoned, emailed, chased and waited. Every one of those contacts cost money. None of them created value.
Speeding up the handling of failure demand is useful. Removing the failure is worth far more.
Now look at Worcestershire County Council.
Its Bus Pass AI Agent checks the documents residents upload with concessionary bus pass applications - flagging an unsuitable photo, a date of birth that doesn’t match, a proof of address that won’t be accepted. A person still makes the final decision.
In its first eight months, almost 2,500 applications were processed with its help and nearly 7,500 documents checked. It caught hundreds of problems early on that would otherwise have delayed applications, before they turned into a rejection, a phone call and a second attempt.
Wiltshire’s AI is helping a stretched team cope with the volume of things that have already gone wrong. Worcestershire’s is stopping some of them going wrong in the first place.
Both are legitimate. Both are well governed. They tackle different services, so this isn’t a like-for-like comparison. But only one of them is designed to reduce next year’s demand. The time Wiltshire saves could go into fixing causes too, but that has to be a deliberate choice.
Most AI implementations start with the tool and then search for a problem to solve. The noisiest problems tend to get the most attention, and that noise is usually coming from downstream in the service: the inbox, the contact centre, the complaints queue.
But the cause of that noise often sits much further upstream. A poorly designed form, a broken handoff, a missing update or information that nobody can easily find.
Starting with the noise might help you deal with it faster. Starting with the cause can reduce the volume in the first place.
That’s where much of TVI’s work with councils is focused. Understanding what’s happening across services, identifying where things are going wrong and finding opportunities to improve them before deciding what technology, if any, is needed.
We look at this through three lenses:
Outcomes: what is actually happening for residents, and where in the journey it goes wrong.
Experiences: what it is like on both sides of the counter: the resident chasing, and the officer answering the same question for the fourth time.
Efficiencies: what each avoidable contact costs, so the case for fixing the cause can be counted, not just felt.
A ten-minute check before your next AI investment goes to Cabinet
One - take your top three complaint categories from last year. For each, can you name the point in the journey where it started going wrong?
Two - look at your current AI pipeline. How many use cases make handling faster, and how many make the need to handle disappear?
Three - if residents’ complaints keep getting longer, does anyone own the question of why they’re writing them at all?
If you work in complaints, customer services or digital in a council, you probably already know which handful of services your complaints keep coming from. If you’re a Chief Executive or Director of Resources, your complaints manager or Head of Customer Services does. Get in touch with Triple Value Impact if you’d like the three-lens view we use with councils to trace complaints back to their cause – outcomes, experiences, efficiencies. No cost, no catch, just the lenses applied to your context.
AI will sit on both sides of the counter whether councils plan for it or not.
The real choice is whether yours is there to manage the queue - or to make it shorter.
