Two federal AI stories broke within a day of each other last week, and together they say a lot about how fast this government actually moves.

On July 29, Senate Commerce Committee Chairman Ted Cruz postponed the committee’s planned markup of a package of AI bills, including provisions meant to preempt the growing patchwork of state AI laws. Republicans and Democrats couldn’t reach agreement in time. The committee will take it up again on September 5, more than five months after the White House released its National AI Legislative Framework calling for exactly this kind of federal action. The next day, July 30, GSA livestreamed a governmentwide showcase proving it had already automated 850,000 hours of employee work, pushed internal AI use above 70 percent, and published a free, reusable playbook explaining how it got there.

Same government, same month, two very different clock speeds. And GSA isn’t the only agency proving the point.

What Actually Stalled in the Senate

The Commerce Committee’s AI package was supposed to do what Senator Blackburn’s TRUMP AMERICA AI Act discussion draft and the White House framework have been pointing toward since March: give agencies, states, and industry a single federal standard instead of a fifty-state compliance maze. The sticking point wasn’t whether to act. Both parties agree the state patchwork is a real problem. It was scope. Democrats pushed amendments to narrow how far federal preemption would reach into state consumer-protection and AI-safety laws. Republicans, backed by the administration, wanted broader preemption. Neither side moved far enough by July 29. Kids’ online safety provisions, including KOSA and related bills, split off and advanced on their own. The core AI guardrails and preemption fight got rescheduled, not resolved.

This is the second time in five months that “the federal AI policy moment” has been announced and then slipped. The discussion draft is still a discussion draft. No comprehensive federal AI statute has passed. Agencies operating under EO 14409, BOD 26-04, and the quantum cryptography orders are working from executive directives with hard compliance dates, not from a statute that settles the underlying preemption question.

What Agencies Did Without Waiting

GSA didn’t need a statute to hit 70 percent internal AI adoption, up from 10 to 15 percent at the start of this administration. It didn’t need one to identify 850,000 of the million work hours it’s targeting for automation this year, or to publish its 37-page Elimination, Optimization and Automation playbook in June, or to sign two dozen agency agreements onto its USAi.gov sandbox with dozens more in progress.

It’s not alone. The Department of Energy told the same GovExec conference in mid-July that two internal AI tools, a generative platform called Joulix and a data platform called Quanta, are on track to deliver an estimated $74 million in annual operational savings. Joulix now serves 21,000 users across DOE’s 88 department elements and all 17 national labs. Quanta adoption grew nearly 5,000 percent in eight months after the Office of Electricity used it to sort a billion documents in twelve minutes against a tight executive order deadline, work that would have taken weeks by hand. At CMS, officials say Copilot usage is running around 80 percent weekly among staff, saving roughly five and a half hours per employee per week, while the agency has also used AI and analytics to help block more than $2 billion in improper payments before they went out the door.

None of that required Congress. It took an internal mandate, a methodology, and the engineering capacity to actually execute it, at three different agencies, in three different mission areas.

A Caveat Worth Taking Seriously

Not everyone is convinced the adoption numbers tell the full story, and the skepticism deserves a place in this conversation rather than a footnote. A recent GAO review of how DOD, DHS, GSA, and VA are acquiring AI capabilities found agencies learning hard lessons in isolation and not sharing them across government. GAO’s own science and technology arm has also found that even strong-performing AI agents complete only about 30 percent of complex tasks autonomously without error, which means the other 70 percent still lands on a person, and that person needs the training and context to catch the mistake before it moves downstream. Faster adoption and real efficiency gains are not automatically the same thing as good judgment at scale, especially in functions like benefits determinations or enforcement actions where the right answer depends on nuance a model can flatten out.

That’s not a reason to wait. It’s a reason to pair adoption speed with real oversight of where AI is making decisions versus where it’s just saving someone time on a first draft.

The Real Lesson: Don’t Wait on the Legislative Clock

For agencies and contractors, the temptation is to treat AI legislation as the starting gun. Wait for the federal framework, wait for preemption clarity, then build. GSA, DOE, and CMS are all the counterargument. The rules that actually govern federal AI adoption right now are executive: agency directives, OMB and GSA guidance, procurement vehicles like OneGov, and internal playbooks like EOA. Those move on an executive timeline, which has proven to be months, not years. The legislative framework, the piece that would give the whole ecosystem including state governments and private industry real long-term certainty, is still moving on a congressional timeline. That timeline just showed it can slip by a season with no real cost to anyone in the room.

That gap matters for planning. If your agency’s AI strategy has “wait for the federal AI law to pass” anywhere in it, that’s not a plan. It’s a placeholder. If your compliance posture depends on federal preemption arriving to simplify a state-by-state patchwork, plan as if that patchwork sticks around through 2027. Two chambers and two separate court tracks are currently working through AI policy on their own schedules, and nothing suggests they’re converging on a single deadline anytime soon.

The Bottom Line

Congress is still negotiating the rules. GSA, DOE, and CMS already built machines that run without them, and produced real savings and real oversight questions along with them. Agencies and contractors who treat AI adoption as an operational problem to solve now, using the executive playbooks that already exist, will be ahead regardless of what the Senate Commerce Committee does on September 5. Anyone still waiting for legislative clarity is, at this point, waiting for a train that’s already been rescheduled twice.

That caveat about the GAO’s findings points to something bigger than any one Senate markup. The real gap in federal AI right now isn’t adoption, it’s accountability.

Government doesn’t have an AI adoption problem anymore. It has an accountability problem.

That’s the real story hiding underneath this year’s flood of agentic AI headlines. A survey of more than 200 federal IT executives released this spring found that 53% of agencies are now exploring or actively piloting agentic AI, with another 15% already implementing it. Only 6% said they weren’t considering it at all. OMB’s own 2025 use case inventory backs this up: reported federal AI use cases more than doubled year over year, crossing 3,000, with HHS, NASA, VA, DOE, and DOJ reporting the largest numbers of individual use cases. Agencies aren’t dabbling anymore. They’re deploying.

Here’s the part that should give every public sector leader pause. 77% of those same federal leaders say oversight frameworks for agentic AI are essential. Fewer than a third have actually built one. Only 20% have defined policies for pre-deployment testing. Only 8% have a real incident response framework. Fewer than a third have a documented “kill switch” procedure for when an autonomous system needs to be stopped mid-action.

That gap, between what agencies know they need and what they’ve actually built, is where the risk lives.

It matters more in government than almost anywhere else, because agentic AI isn’t just answering questions anymore. It’s initiating actions: triggering workflows, touching case files, moving money, flagging people for review. In benefits administration, fraud detection, emergency response, and public safety, an AI agent’s output isn’t a suggestion sitting in a chat window. It’s a decision with a person on the other end of it. When something goes wrong, “the model made an error” isn’t an acceptable answer. Someone has to be able to trace exactly what data the system touched, what permissions it used, and why it acted the way it did. Brookings has documented how “black box” decision-making persists even as adoption accelerates. Right now, most agencies can’t consistently explain their own systems.

This isn’t an argument for slowing down. The momentum behind agentic AI in government is real, it’s bipartisan, and it isn’t reversing. Leaders see genuine upside here: faster service delivery, more responsive case handling, staff freed up from repetitive work at a time when federal headcount is shrinking, not growing. Slowing adoption to a crawl while governance catches up isn’t realistic, and honestly, it isn’t even the right goal.

The right goal is building the accountability layer at the same speed as the capability layer, instead of treating it like a phase-two problem.

In practice, that means a few unglamorous things happening alongside every new pilot, not after it. Logging and audit trails for every action an agent takes, not just its outputs. Human-in-the-loop checkpoints scaled to risk: tight, mandatory human approval for anything touching national security, critical infrastructure, or benefits decisions, and a lighter touch for the low-stakes stuff. Vendor contracts that actually assign liability instead of leaving it vague. And a real, tested kill switch, not a policy document that just describes one.

The good news is agencies already know how to prioritize this correctly. In that same survey, nearly 90% of respondents said they require logging and audit trails for all agentic actions, and nearly 80% mandate human-in-the-loop review for high-risk work. The instinct is right. What’s missing is implementation, and implementation is a solvable problem, not a philosophical one.

The agencies that get this right won’t be the ones that adopted agentic AI fastest. They’ll be the ones that can explain, a year from now, exactly how every consequential decision their systems made actually got made. That’s a much higher bar than a successful pilot demo, and it’s the one that will actually matter when something breaks in production, in front of a citizen, an inspector general, or a congressional committee. As one FedScoop commentary on the topic put it, government is no longer just setting the rules for AI. It’s becoming one of its most active operators.

That’s a bigger shift than most of the adoption headlines let on, and it deserves to be treated like one.

This is the gap we spend our days closing at Elevate Government Solutions: deploying and integrating agentic AI in a way that treats logging, human-in-the-loop review, and accountability as part of the build, not an afterthought bolted on after the pilot. If your team is wrestling with how to move from pilot to production without losing control of the process, I’d welcome the conversation.