Public services in multiple countries have seen sharp increases in complaints, petitions and benefit applications since 2022, a trend researcher Chris Schmitz calls “agentic flooding.” Schmitz’s analysis finds similar patterns across different systems, but it stops short of asserting direct causation between AI tools and the rise.
Scope and figures
The paper examines 84 cases across 11 jurisdictions. In the United Kingdom, complaints to the housing ombudsman rose from 2,600 in 2022 to just over 7,000 last year. The United States Consumer Financial Protection Bureau (CFPB) recorded approximately five times as many complaints over the same period. Schmitz also reports comparable jumps in judicial petitions in Brazil and parliamentary petitions in Germany.
Patterns, causes and challenges
Schmitz reports that most cases follow a common arc: submissions were roughly flat before 2022 and then climbed as certain AI tools became more widely available. For methodological reasons, the paper does not claim AI is the direct cause, but it notes that the growth has generally not slowed, suggesting continued upward pressure.
One mechanism identified is the reduction of administrative burden: AI assistants can help people gather documents, compose filings or convert photos of letters into usable text. Schmitz argues this lowers the effort required to submit claims that previously would have been abandoned. He contrasts this with last year’s influx of automated, low-value bug-bounty reports, but emphasizes a key difference: “The vast majority of cases we find are people who are entitled to claim for something, claiming for that thing,” he says.
Schmitz also points to the operational strain on public services. If agencies must process substantially larger volumes with existing budgets, resource pressures could increase. He suggests the rise presents an opportunity to redesign online interactions for an era in which AI assistants are widely used: “A big part of making AI go well is being able to detail out what the good version of things looks like,” he says.
Schmitz’s paper will be presented next month at the AI Ethics and Society conference and includes the dataset underlying the 84 cases across 11 jurisdictions.
Original source: TechCrunch AI