Flood warning: How citizens’ AI agents will swamp public services | Computer Weekly

Creating friction

Many public services rely on friction to stay viable. They depend on slow, confusing, frustrating user experiences to put off those otherwise eligible – how often do people just get fed up trying, and give up? This is both unfair and politically convenient. You could say “’twas ever thus” – until now.

From parents seeking special needs support to property owners appealing council tax bands, it’s often the friction of bad service design that restrains demand, not the law.

AI – specifically AI agents – will remove that friction. Your AI agent will be doggedly relentless in how it accesses public services on your behalf, however byzantine those services may be. It will make sure your application is perfectly crafted to maximise your chances of getting what you want, treating any appeals process as just another stage to be navigated.

Ask your agent

I’m lucky to have played a lot with AI agents recently – the likes of OpenClaw, PicoClaw and Claude Cowork. I recently ran an experiment with OpenClaw – what would an AI agent do if I asked it to tell me whether the council tax band for my house was fair, in comparison to my neighbours?

It came back immediately to tell me the band was higher than all my neighbours and suggested some next steps it could take. At this point I stopped it, as I’d realised something stark.

If I’d have let it, my AI agent would happily have run off to compare neighbours’ floor areas by querying the Gov.uk Energy Performance Certificate API; it would have measured neighbours’ extensions from Ordnance Survey; downloaded Land Registry’s historic house price dataset; searched property websites for number of bedrooms; and researched how best to craft an appeal over my council tax band to the Valuation Office Agency. It would then have written a far better appeal letter than I ever could and submitted it on my behalf. Just like that – all without any intervention on my part.

Now I might still have lost the appeal, but the cost to me in time and hassle would have been negligible compared to even three months ago – one click and about 12p, which is the most expensive it’ll ever be. The friction that stops people from appealing their council tax band just disappeared. Ditto every other public service.

Now what?

Agentic flooding

Welcome to the new frontier of “agentic flooding”, a term coined by Chris Schmitz, a PhD student at Berlin’s Centre for Digital Governance. He’s created a dashboard highlighting increased demand for public services which might be attributed to citizens’ use of AI.

For example, benefit appeals to the Department for Work and Pensions have increased by over 60% since the first benefit-specific AI tools appeared in 2022. And this was before AI agents appeared on the scene at the end of 2025.

Governments are not remotely ready for the coming explosion in demand for their services driven by AI agents. It might take a couple of years, but it’s coming.

Much of this demand will be entirely legitimate. Some of it doubtless will be fraudulent. But demand is demand, and AI agents don’t ever get bored – they negate the friction that used to keep demand in check.

Adding more friction to restrict AI agents would see the government kicking off an AI arms race against its citizens in which both sides lose. Instead, governments will need to clarify – if not tighten – countless rules, policies, processes and regulations, otherwise public services risk being swamped. Along the way, policy grey areas will be eliminated, and that’ll be a loss.

All this won’t be popular, particularly if done in a hurry in response to a crisis.

There’s been a great emphasis in Whitehall on using AI to write better policy papers. I hope they’re using AI to explore the myriad tricky policy responses required to respond to the imminent explosion in demand.

Oh, and by the way – the same will apply for the private sector.

Tom Loosemore is a partner at consultancy Public Digital.