As artificial intelligence simplifies form completion and complaint filing, public services worldwide are experiencing significant increases in the number of applications and requests. In the UK, housing ombudsman complaints more than doubled following ChatGPT’s introduction, jumping from 2,600 in 2022 to just over 7,000 the next year. Similarly, the US Consumer Financial Protection Bureau saw complaints multiply by five times in the same timeframe, with analogous rises noted in Brazilian judicial petitions and German parliamentary requests. Researcher Chris Schmitz has been monitoring this surge, describing the phenomenon as “agentic flooding” in his upcoming paper for the AI Ethics and Society conference. Examining 84 instances across 11 jurisdictions, he identifies widespread evidence that AI tools are transforming how the public interacts with government services.
The challenge now lies in determining appropriate responses. While some new filings appear adversarial, others come from genuine applicants utilizing AI to submit claims they might have otherwise abandoned. Although some view the influx as AI-generated spam, Schmitz interprets it as a unique opportunity to redesign social services for the AI age. His research spans services like welfare applications and judicial appeals, all of which offer online access and can be navigated by AI assistants. Although his study doesn’t conclusively establish AI as the direct cause of increased filings, most cases show relatively stable submissions before 2022 followed by consistent growth correlating with AI’s spread—growth that shows no sign of plateauing, suggesting a continuing upward trajectory.
Schmitz explains that the rising proficiency and accessibility of AI naturally lead to increased engagement: “People are discovering this is something they can do, and it’s becoming progressively easier. What once required detailed context assembly and precise prompting has evolved into simple actions like pasting or photographing a letter and getting strong responses instantly.” A comparable surge occurred in bug-bounty platforms last year, as companies were inundated with low-quality AI-generated security reports that, despite lacking critical issues, still demanded resource-intensive review. Public services face a parallel challenge—handling potentially fivefold more applicants with unchanged resources. However, unlike those bug-bounty submissions often dismissed as worthless, Schmitz observes that most new public service applications come from legitimate claimants entitled to receive aid.
Many deserving applicants may previously have been deterred by the administrative burden of applying, a well-known barrier in policy circles. AI’s ability to alleviate this burden presents an opportunity to restructure these services to be more AI-friendly. Although it is a considerable undertaking yet to be adequately addressed, Schmitz remains hopeful. He highlights that effective AI integration depends on clearly envisioning optimal outcomes, noting, “Anyone who’s used ChatGPT to help with a tax return knows there is a beneficial way to be assisted.” This juncture could serve as a critical moment to reevaluate and fundamentally redesign the entire process to better leverage AI’s potential in public service delivery.