When Forrester wrote about the creation of Google Public Sector in 2022, the question was whether Google could build the industry credentials to compete more seriously for government business. Four years later, the conversation has changed.

At the Google Public Sector Summit 2026 on October 20 in Washington, DC, there wasn’t one major announcement that defined the day. Instead, federal, state, local, and education leaders showcased how they are applying AI to government missions. What struck us most was how different the conversation felt from even early 2025, let alone 2022. The focus has moved beyond one-off pilots toward workforce access at scale and moonshot problems where process inefficiencies or skill shortages are forcing change.

From AI Pilots To “Super Intelligence” And Moonshots

The use cases shared by Google Public Sector’s clients reflected that ambition. The Department of War discussed putting Gemini into the hands of more than 3 million personnel. The Army showed how AI could make fragmented logistics data accessible. Iowa discussed using Gemini and its data cloud for child-welfare case histories. At the US Patent and Trademark Office (USPTO), thousands of highly specialized patent examiners face a roughly 700,000-application backlog. But despite the moonshot language, the conversations were remarkably pragmatic: How can technology help agencies accomplish missions that existing people, processes, and systems struggle to address?

Fraud detection was naturally top of mind. The White House CIO discussed the opportunity to apply technology to fraud, while other conversations explored frontline services and fraud detection across government. That’s little surprise given Executive Order 14249, “Protecting America’s Bank Account Against Fraud, Waste, and Abuse,” signed March 25, 2025, which elevated fraud prevention and better use of government data as administration priorities.

Moving Fast Changes The Risk Equation

Security didn’t disappear amid the focus on speed. In fact, it became inseparable from it. A panel featuring Justin Fanelli, CTO of the US Navy; James Wolff, CIO of the US DOE’s National Nuclear Security Administration; Vice Admiral (Ret.) TJ White, chief of Texas Cyber Command; and Thomas Browning of Rhombus Power wrestled with what security means when technology and threats are changing faster than traditional processes can accommodate.

Both the private and public sector are overwhelmed by the number of vulnerabilities discovered and even more so by how quickly exploitation of those vulnerabilities occurs. Agencies still need to secure data, endpoints, infrastructure, and, increasingly, agents. It is not humanly possible to keep up — so much of the discussion focused on “human on the loop” (or “humans in the lead”) rather than “human in the loop” for risk management specifically.

What Stuck With Us

The summit likely left attendees with two competing impressions. AI is now showing that it can create genuine new possibilities for government, but at the same time, many of the hardest questions about cost, risk, skills, and technology-led change feel strikingly familiar, especially when:

  • Gemini is used to optimize Gemini. Google showed early experimentation using Gemini to recommend model-routing and architecture choices, forecast spend, reuse agent skills, and apply runtime spend caps. It was an interesting glimpse of FinOps shifting further left, with AI helping optimize the cost and efficiency of AI itself.
  • Barry Schindler’s career-spanning tools stole the show. The USPTO’s acting commissioner for patents came armed with props, tracing technology in his lifetime from the slide ruler to the calculator to AI. Each leap felt a little like cheating, he joked, but you’d be a fool not to take advantage of it. The catch: You can’t create a workforce that needs the tool to do the basics of the job. Another panelist made a similar observation about vibe coding, which quickly separated the strongest talent from the weakest. AI can amplify expertise, but it can’t replace the expertise needed to know when the answer is wrong.

That’s an important tension for government as it embraces AI. Rather than just make every task faster, the opportunity is to use AI to amplify expertise without inadvertently eliminating the expertise that makes the technology useful in the first place.

If you want to talk more on public-sector AI trends, please reach out and schedule an inquiry with our lead, Sam Higgins, or myself.

Share