To Rethink Government, DOGE Must Unshackle AI within Government

N ext year, the U.S. government alone will likely spend more than the GDP of France and the U.K. combined. It will also likely pay people not entitled to receive it more money than the GDP of Sweden, and, if the recent past is any guide, do very little to recover most of these funds. We shouldn’t tolerate such mismanagement, and yet we do. However, due to signals sent by President-elect Trump and the incoming Congress — best illustrated by the Department of Government Efficiency (DOGE) — there’s hope that things might change for the better.

Some of the world’s best engineers, armed with decidedly non-governmental ways of thinking, are filtering into Washington to respond to the challenge of making the government more efficient. This may be an unprecedented opportunity to revolutionize the federal bureaucracy through technological innovation. First, though, the reformers must lay the proper groundwork within government.

The potential is clear. Watchdogs like the Government Accountability Office and various agencies’ inspectors general have identified hundreds of billions in potential savings through better management and regulatory streamlining — suggestions that Congress often ignores. Elon Musk and other tech visionaries, along with entrepreneur Vivek Ramaswamy, possess different tools. Through artificial intelligence and advanced automation, they can transform government operations at a far deeper level.

Consider how SpaceX revolutionized space-launch costs, or how Tesla reimagined electric vehicle production. Similar innovations could be applied to eliminate improper payments ($236 billion a year, plus more that we don’t know about), detect fraud in real time (right now running between $233 billion and $521 billion annually), and prevent duplicative spending before it occurs. Rather than simply identifying fraud after the fact (at least $100 billion for Medicaid and Medicare alone), artificial intelligence systems could proactively optimize resource allocation across agencies, spotting patterns and inefficiencies that evade human auditors.

Evidence for such transformation is already emerging. In 2023, Google Cloud demonstrated the remarkable potential of AI in financial-crime monitoring at institutions like HSBC, where AI scanning reduced false positives by 60 percent while enabling two to four times the number of identified financial crimes. 

The scale of federal transactions makes traditional human-led “pay and chase” compliance increasingly obsolete. Modern AI systems can scan billions of transactions in real time, preemptively identifying patterns and anomalies that would take human investigators months or years to discover. It’s a transformation that could save taxpayers hundreds of billions annually.

However, unleashing these valuable AI tools requires more than technological capability. Just as SpaceX’s revolution was enabled by the 2004 deregulation of commercial spaceflight, unleashing AI’s full reform potential requires tackling the complex web of regulations binding federal IT.

Our IT systems face a daunting array of overlapping requirements, including OMB’s AI impact assessments, federal cybersecurity standards, and countless agency-specific rules. The challenge becomes even more complex when considering the landscape of informal guidance documents that often carry the weight of regulation in practice.

Consider the case of machine-translation AI. While the federal government technically permits its use, implementers must navigate requirements from multiple agencies, including the Office of Management and Budget, the General Services Administration, and the Department of Justice’s civil rights division. Similar regulatory tickets and overlapping guidelines related to welfare, non-discrimination, and other concerns will likely suppress fraud-detection AI and deter decisive engineering solutions.

Somewhat ironically, the solution lies in using AI to clear this regulatory underbrush. Ohio’s recent success with Deloitte’s RegExplorer AI tool, which eliminated 2.2 million unnecessary words from that state’s administrative code, shows what’s possible. Similar technology could help identify and consolidate the universe of formal regulations and informal guidance that currently hamper federal modernization efforts. Since many informal guidance changes don’t require formal administrative procedures, this could offer quick — yet impactful — wins to further DOGE’s ambitious goals.

The Federal Acquisition Regulations that govern defense and civilian agency purchases are an important first target. Containing thousands of clauses, these rules are onerous, duplicative, and sometimes even contradictory. The result is waste, and the use of outdated technology due to lengthy procurement processes. Simplifying this rulebook with AI will yield across-the-board regulatory relief that will enable agencies to both cut considerable procurement costs and acquire the state-of-the-art tech needed to become more efficient.

Next, the incoming administration should implement a “cash for cuts” system, leveraging existing resources like the Technology Modernization Fund (TMF), an IT fund created under the first Trump administration. Agencies applying for TMF funds would need to identify regulations and guidance for removal, and would in return receive both regulatory freedom and fiscal resources to focus on coding rather than compliance.

Augmenting AI are other emerging technologies. Blockchain could provide unprecedented spending transparency, while advanced data analytics could map and consolidate program overlaps. Such waste-busting solutions go far beyond traditional oversight.

With the groundwork laid, DOGE could serve as an innovation hub, bringing Silicon Valley–style disruption to federal operations. Instead of just implementing watchdog recommendations, transformative technologies could fundamentally restructure how government works, potentially saving trillions while cutting down bureaucracy and improving services.

The question isn’t just one of making government more efficient within its current paradigm; it’s how we can reimagine government operations for the age of artificial intelligence. Of course, our government will not be truly efficient until Congress finally reforms entitlement programs, devolves functions to the states, and cuts out cronyism.

The path forward requires bold leadership willing to challenge established bureaucratic processes, reform regulations, and embrace technological innovation. With the right combination, we could achieve what generations of government reformers have only dreamed of: a more efficient, responsive, and accountable federal government.

Veronique de Rugy is the George Gibbs Chair in Political Economy and a senior research fellow at the Mercatus Center at George Mason University. Matthew Mittelsteadt is a research fellow and technologist with Mercatus’ AI and Progress Project.