Britain’s Bureaucracy Gets More Paperwork As AI Advances

🔍 Read the full analysis: Britain’s Bureaucracy Gets More Paperwork As AI Advances on ThorstenMeyerAI.com

A Sky News report published October 20 found that written complaints, Freedom of Information requests and employment tribunal claims have risen across multiple UK public services since ChatGPT launched in November 2022. The trend is putting pressure on services that rely on people writing letters and forms, although the report does not establish that AI caused the increase or that the additional submissions are mostly invalid.

Data examined by Sky News showed increases since ChatGPT launched in several complaint channels: 144% for household complaints about water companies, 78% for complaints to English local authorities and 71% for written complaints to GP practices. Complaints to the Department for Work and Pensions rose 64%, while individual employment tribunal claims increased 62%. These figures describe growth over the period cited in the report; the source does not provide a common baseline count for each channel in the supplied material.

Other reported increases included 49% for complaints about legal services, 35% for written complaints to the NHS, 34% for complaints to police forces and 23% for complaints about public services. Sky News said the lines for these channels had generally been relatively flat before late 2022 and then rose. Each service may also have its own reasons for the change: the report, for example, noted difficult recent years for water companies.

Sky News technology correspondent Rowland Manthorpe also tested 100 recent FOI requests with an AI-text detector. It labelled 89% fully AI-written and 9% human, while a smaller sample of requests from 2019 was labelled entirely human. Manthorpe cautioned that AI detectors are unreliable and that the small test cannot prove AI caused the wider increase. Separately, a survey of civil servants cited by Sky News found that 58% said AI-generated public submissions significantly affected their work.

Pressure on Paperwork-Based Services

The reported rise matters because many public services still depend on people setting out a case in writing, then on staff reading and responding to it. AI can lower the effort needed to prepare a complaint or claim, potentially bringing more people into processes they previously found difficult to use. It can also increase the volume and complexity of material that public bodies must handle.

That creates a difficult policy choice. Restricting submissions or adding fees could reduce workload, but it may also deter people with valid complaints. The report cites the Financial Ombudsman Service as an example: it found about a third of its complaints were AI-generated, began charging some professional representatives to bring complaints, and later saw complaint numbers fall. That change does not by itself show whether the remaining complaints were more justified or whether some people were deterred from seeking redress.

Evidence cited in the report suggests the new volume cannot simply be treated as spam. At the DWP, the proportion of complaints upheld has fallen from the 50% range to close to 40%, according to Sky News. Yet a lower success rate does not mean fewer valid complaints in absolute terms when the total number has grown substantially. Researcher Chris Schmitz, whose work the report discusses, said most of the cases he identified involved people claiming something to which they were entitled.

From Chatbot Drafts to Automated Claims

The pattern described by Sky News extends beyond Britain. A paper by researcher Chris Schmitz, reported by TechCrunch in September, documented 84 possible cases of what he calls “agentic flooding” across 11 jurisdictions. The source material says most were roughly flat before 2022 and rose afterward. The examples include the UK Housing Ombudsman, where complaints grew from about 2,600 in 2022 to just over 7,000 last year, and the US Consumer Financial Protection Bureau, where complaints increased fivefold.

In May, the Information Commissioner’s Office issued guidance on AI-generated FOI requests after public authorities reported higher volumes and more complicated submissions. The ICO said an AI-written request remains valid, authorities do not need to determine who or what wrote it, and they can seek clarification when a request is poorly formed or unclear. That guidance addresses handling requests, not the broader question of how public services should adapt their application and appeals systems.

For now, many people use chatbots to draft text that they then submit themselves. The next change could be AI agents that submit or pursue claims on a person’s behalf. Manthorpe’s report demonstrates an agent service from Meta and raises examples such as automated parking appeals and Delay Repay claims for delayed rail journeys. These are scenarios, not evidence that such systems are already generating mass claims in the UK.

““It’s like being bombarded with correspondence at speed.””

— Civil servants surveyed in research cited by Sky News

What the Trend Cannot Prove

The reported timing and broad rise across several channels do not establish that ChatGPT or other AI tools caused the increases. The figures cover services with different rules and circumstances, and the supplied source material does not give full counts, consistent measurement periods or a comparison group that would isolate AI’s effect. Other changes in public services, complaint practices or underlying problems could also affect volumes.

The detector test is limited evidence: it involved 100 recent FOI requests, a smaller 2019 sample and a tool Sky News’ correspondent acknowledged may be unreliable. It cannot reliably determine the authorship of all submissions or show how much AI contributed to the wider trend. The survey figure of 58% records civil servants’ reported experience; it does not measure the number of AI-written submissions across government.

It is also unclear how quickly people will move from using AI to draft a submission to using agents that file and follow claims automatically, or how many such claims would be valid. The report does not provide a detailed estimate of the staffing or financial costs to public bodies, nor a settled government plan for responding.

Redesigning Forms and Appeals

The immediate response from public authorities is likely to involve handling the requests they receive under existing rules. For FOI requests, the ICO’s May guidance says authorities can ask for clarification where needed but should not reject a request simply because it was written by AI. The broader pressure may prompt agencies to review how they receive, triage and process complaints and claims.

Loosemore’s proposed alternative is to replace essay-style applications with structured data, so people provide information through clearer, more direct processes rather than composing long narratives. He warned that adding barriers to submissions would create a contest with technology providers, which could find ways around those restrictions. His comments are a policy argument, not an announced government programme.

Further evidence will be needed to show whether the increases continue, how much AI contributes and what share of additional cases lead to a successful outcome. The next major test may come as automated agents make it easier to submit claims at scale. Until public bodies publish more detailed data or announce service changes, the balance between managing workloads and keeping valid routes to redress open remains unresolved.

Key Questions

What did the Sky News report find?

It reported that written complaints and claims rose across several UK public services after ChatGPT launched in November 2022. The increases varied by channel, including 144% for household complaints about water companies and 64% for complaints to the DWP over the period cited.

Does the report prove AI caused the increase?

No. The report describes a rise that began after ChatGPT launched and presents evidence consistent with increased AI use, but it does not establish causation. The AI-text detector test was small, and the correspondent noted that such detectors are unreliable.

Are the additional complaints mostly invalid?

The report does not support treating the increase as mostly invalid. It says the DWP’s share of complaints upheld fell from the 50% range to close to 40%, while researcher Chris Schmitz said most cases in his research involved people claiming something they were entitled to. Those findings do not establish the validity of every new submission.

What does the ICO say about AI-written FOI requests?

The ICO’s May guidance says AI-written FOI requests remain valid. Public authorities do not have to determine who or what wrote a request, and they can ask for clarification if it is unclear or poorly formed.

What could change as AI agents become more common?

AI agents may be able to submit or pursue claims on a person’s behalf, beyond simply drafting text. The report raises automated parking appeals and rail delay claims as examples, but does not establish how widely agents are currently being used for those purposes.

Source: ThorstenMeyerAI.com