Code for America chief executive Amanda Renteria argues that the US needs a national model for using AI to tackle public problems, with safeguards and delivery capacity built in. Her proposal offers a practical test for government AI: does it improve people’s lives, and can people challenge a wrong decision?
Watch Desk analysis
What happened
In a commentary published by FedScoop on 26 October, Renteria says America.gov is a useful starting point for finding official information, but is still principally a chatbot. She argues that better public services need more than a promising tool: government must connect systems, improve data, redesign workflows, train staff and maintain services after launch.
Renteria sets out four principles: keep people informed and in control, with a real way to challenge AI-influenced decisions; build shared infrastructure and publish evidence; bring technologists, public servants and communities into the work; and invest in delivery. She says systems should be tested for accuracy, bias, privacy and security both before and after reaching the public.
She points to Code for America’s GetCalFresh work, which she says cut California’s food-assistance application from more than an hour to roughly 10 minutes. The organisation is also building a SNAP Policy Navigator with Anthropic to help caseworkers interpret complex policy.
Why it matters
The proposal shifts attention from AI pilots to what happens when a service meets the person trying to use it. Renteria argues that useful technology depends on the less glamorous work of joining up systems and supporting the staff who use them. The ribbon-cutting gets the photograph; the hand-off between departments gets the job done.
Her examples also show why outcomes matter more than a dashboard turning green. Renteria says governments should ask whether errors fell, public servants gained useful time and people could challenge incorrect decisions. Those are concrete measures, even if the commentary does not establish that a national model has been adopted or that AI alone produced GetCalFresh’s improvement.
Our read
This is a useful policy argument because it pairs safeguards with the work needed to make services function. The strongest test is not whether government can launch an AI tool, but whether it can keep that tool accountable and improve the service around it. Start with the public goal, measure the result, and keep a human route to challenge a bad decision. It sounds less futuristic than a chatbot; it is considerably more useful.
What to watch
- Whether US agencies adopt shared protections, infrastructure and public evaluation methods.
- What the SNAP Policy Navigator does in practice, and how caseworkers can question its recommendations.
- Whether government AI projects publish evidence on errors, service outcomes and people’s ability to challenge decisions.
Discussion spark: Should governments be required to show that an AI-supported service improves outcomes and offers a meaningful way to challenge mistakes before expanding it?
Sources and evidence
- Source update (26 October 2026, 15:16 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.