AI Agents Are Rewriting the Public-Service Queue

Generative AI is lowering the cost of dealing with government, and public agencies are beginning to feel the consequences in their inboxes. TechCrunch reports that complaints to the U.K. housing ombudsman rose from about 2,600 in 2022 to just over 7,000 the following year. The U.S. Consumer Financial Protection Bureau saw complaints increase fivefold over the same period. Brazilian judicial petitions and German parliamentary petitions showed comparable jumps.

Researcher Chris Schmitz calls the broader pattern “agentic flooding.” In a paper scheduled for presentation at the AI Ethics and Society conference, he examines 84 potential cases across 11 jurisdictions. For methodological reasons, the study does not claim that AI directly caused every increase. Still, most cases share a recognizable pattern: volumes were broadly flat before 2022, then accelerated as AI tools became more widely available. In many cases, the growth has not yet slowed.

Key points

• AI is reducing administrative burden. Applying for benefits, filing a complaint or preparing an appeal often requires collecting documents, understanding specialized rules and producing persuasive written explanations. AI can now turn a photographed letter or a pasted notice into a usable draft with far less effort.
• More submissions do not necessarily mean more spam. Schmitz distinguishes public-service requests from the low-quality reports that have overwhelmed some bug-bounty programs. His finding is that the vast majority of cases involve real people seeking something they are entitled to claim.
• Capacity will become a serious issue. Even legitimate applications consume staff time. Agencies must review claims, verify evidence, detect abuse and make decisions, often without a matching increase in budgets or personnel.
• The increase may expose hidden demand. Some people may not have applied in the past because the process was too difficult, time-consuming or intimidating. AI is making that previously invisible demand visible.

Why it matters

The important change is not simply that AI can write more forms. It is that public-service statistics have traditionally measured submitted applications, not the people who abandoned the process before submitting anything. Barriers involving language, legal knowledge, time and confidence can suppress demand. When an AI assistant removes part of that burden, institutions may discover that their apparent workload was never a complete measure of public need.

That discovery creates both an opportunity and a risk. Agencies could redesign their online services around clearer rules, structured evidence and secure AI assistance. An assistant might help a resident understand eligibility, organize documents or prepare a first draft, while human officials retain responsibility for verification and final decisions. Such systems would also need strong privacy protections, audit trails and safeguards against confidently generated errors.

Simply adding more reviewers may not be enough if request volumes continue to rise. Nor would labeling every AI-assisted submission as abuse be appropriate when many applicants have valid claims. The better response is to separate the legitimate increase in access from manipulation and low-quality automation, then build processes capable of handling both.

“Agentic flooding” is therefore more than a warning about overloaded government inboxes. It is evidence that administrative friction has long prevented some people from using services available to them. AI may turn that friction into a public-sector capacity crisis—but it may also provide the impetus to redesign public services for a world in which assistance is always available.

Source: TechCrunch AI