The Great Reinvention logo THE GREAT
REINVENTION

Sunlight as Savings: AI-Assisted Public Audits of Every Federal Department

Proposed legislation: The Federal Transparency and Continuous Audit Act

Public Audits and Transparency: Cost-Benefit Analysis

The federal government spends roughly $7 trillion a year, yet large parts of it are effectively unaudited in any rigorous, continuous, public way. The Department of Defense has never passed a clean, comprehensive financial audit. Improper payments run well over $160 billion annually. The Government Accountability Office (GAO) publishes an annual catalog of duplicative and overlapping programs that, year after year, identifies tens of billions in unrealized savings. The common thread is not that no one is looking — GAO, agency Inspectors General, and the Office of Management and Budget all look — but that auditing is episodic, backward-looking, fragmented across agencies, and often invisible to the public who funds it.

The proposition behind this proposal is to make auditing continuous, comprehensive, AI-assisted, and transparent: deploy modern data-analytics tools to examine spending across every federal department in something closer to real time, surface waste, fraud, and duplication systematically, and publish the findings where citizens can see them. The target is $100 billion in annual savings. This analysis examines where that money would come from, whether the figure is realistic, and what genuine continuous auditing would require.

Where the Savings Come From

Improper Payments and Fraud

The foundation of the savings case is the same well-documented problem that underlies several federal-integrity reforms. GAO reported an estimated $162 billion in improper payments in fiscal year 2024 (about $186 billion in FY2025), and separately estimates annual fraud losses of $233 billion to $521 billion. Cumulative improper payments since FY2003 total roughly $2.8 trillion. Continuous, analytics-driven auditing attacks this directly: rather than sampling payments after the fact, it monitors transaction streams for the anomalies, duplicates, and eligibility mismatches that signal error or fraud.

The promise of AI-assisted auditing is precisely that it can examine the full population of transactions rather than a sample. Traditional financial audits review a fraction of records and extrapolate; machine analysis can flag every suspicious record across millions of transactions, dramatically raising detection rates. Treasury's experience with AI-enhanced payment-integrity tools — which it has credited with billions in prevented and recovered improper payments — is a concrete demonstration of the approach working at federal scale.

Duplication, Overlap, and Fragmentation

The second pillar is GAO's own long-running duplication work. GAO's annual reports on fragmentation, overlap, and duplication have, since 2011, identified opportunities that Congress and agencies have used to achieve roughly $774 billion in cumulative savings and revenue increases — and GAO's 2026 annual report flags the potential for an additional $100 billion or more in future financial benefits from acting on its open recommendations. Earlier reporting found that more than $400 billion was being spent annually across roughly 1,500 duplicative, fragmented, or inefficient programs. Continuous auditing institutionalizes this hunt rather than leaving it to an annual report that agencies are free to ignore.

Recovered Funds and Deterrence

Auditing produces savings in two ways: direct recovery of misspent funds and the deterrent effect of knowing that spending is continuously monitored and publicly reported. The deterrent value is real but inherently hard to quantify; transparency tends to discipline spending behavior even where no specific recovery is recorded.

Projected Figures and the Realistic Range

The proposal targets $100 billion per year. The honest assessment has two parts.

On one hand, the figure is grounded in real numbers. GAO's own work points to $100 billion-plus in achievable savings from duplication alone, improper payments exceed $160 billion annually, and fraud losses are estimated in the hundreds of billions. The raw opportunity comfortably exceeds $100 billion.

On the other hand, there is a crucial distinction between identifying savings and capturing them. GAO has identified vast savings for over a decade, yet much of it remains unrealized because Congress and agencies must act on the findings — and frequently do not. Auditing surfaces waste; it does not, by itself, cut it. The $100 billion figure is therefore realistic as an identification target and plausible as a capture target only if paired with mechanisms that force action on what audits reveal. A continuous-audit program that produces brilliant findings nobody acts on would save far less.

Additionally, some "savings" in this space are one-time recoveries or prevented payments rather than recurring budget reductions, and not all improper payments represent permanently lost dollars. A fair projection counts recurring structural savings (eliminating duplicative programs, preventing recurring improper payments) separately from one-time recoveries.

The balanced conclusion: $100 billion in annual savings is achievable but conditional — it requires not just better auditing but binding follow-through. The auditing itself is comparatively cheap; the political will to act on it is the scarce resource. Treated as an identification-and-action program rather than an audit-and-file program, the target is defensible.

Mechanism: How Continuous Public Auditing Would Work

The proposed Federal Transparency and Continuous Audit Act would establish three mechanisms. First, it would create a government-wide continuous-auditing capability — an analytics platform, operated with appropriate independence (likely in coordination with GAO and the Inspectors General), that ingests spending and payment data from every department and applies AI-driven anomaly and duplication detection on an ongoing basis rather than annually.

Second, it would mandate public transparency: audit findings, by department, published in accessible form (building on existing infrastructure like USAspending.gov), so that citizens, journalists, and watchdogs can see where money goes and where waste is found. Sunlight is itself an enforcement tool.

Third — and most important for actually capturing savings — it would create an action-forcing mechanism: when an audit identifies duplicative programs or recurring improper payments, agencies would be required to respond within a set timeframe with either a remediation plan or a public justification, and persistent failures to act would trigger budgetary consequences. This is the element that distinguishes the proposal from the status quo, in which GAO's findings can be acknowledged and then shelved.

Administrative and Implementation Considerations

The auditing infrastructure itself is relatively inexpensive compared with the spending it monitors — a favorable cost-benefit ratio, since even modest analytics investment can examine billions in transactions. The principal implementation challenges are data and independence.

Data quality and interoperability are foundational: continuous auditing requires standardized, accessible spending data across agencies, and federal financial data has historically been fragmented and inconsistent. The DATA Act of 2014 made progress toward standardizing federal spending data, and continuous auditing would build on and extend that foundation. Investment in clean, connected, machine-readable data is a prerequisite.

Independence matters because auditing is only credible if it is insulated from the agencies it examines. GAO's independence from the executive branch is precisely what gives its findings weight; a continuous-audit capability must be similarly protected from pressure to soften findings. Coordination with existing Inspectors General avoids duplicating the very oversight apparatus the proposal seeks to strengthen.

Finally, AI-assisted auditing requires human judgment in the loop. Automated systems flag anomalies; trained auditors must distinguish genuine waste and fraud from false positives. The goal is to multiply auditor effectiveness, not replace professional judgment.

International Comparisons and Precedent

Continuous and data-driven auditing has precedents abroad. Several national audit institutions have adopted data analytics to expand audit coverage. Some governments and large organizations have moved toward "continuous auditing" models in which transactions are monitored in near-real time rather than reviewed after year-end. In the United States, the GAO and Inspector General community already represent a sophisticated audit apparatus; the proposal is less about importing a foreign model than about modernizing and making continuous what the U.S. already does episodically. The DATA Act and USAspending.gov are domestic precedents demonstrating that government-wide spending transparency is technically feasible.

Comparison to the Status Quo and Alternatives

The status quo is episodic, fragmented, and frequently ignored. GAO produces excellent annual reports identifying duplication and waste; agency Inspectors General produce thousands of recommendations; OMB tracks improper payments. Yet the same problems recur because findings lack binding force and because auditing is backward-looking and incomplete. The result is a system that diagnoses waste well but treats it poorly.

The main alternative is to strengthen and better fund the existing audit institutions without building a new continuous-auditing layer — a reasonable, lower-disruption path that leverages GAO and the IGs. The continuous-audit proposal goes further by making monitoring real-time and public and by adding action-forcing teeth. A second alternative is to focus narrowly on the largest improper-payment programs rather than auditing every department; this concentrates effort where the money is, but misses the deterrent and duplication-detection value of comprehensive coverage.

Risks, Trade-offs, and Counterarguments

The strongest counterargument is that auditing identifies savings but rarely captures them, because capture requires legislative and administrative action that the political system resists. Critics can fairly note that GAO has identified $100 billion-plus in savings for years with much left on the table. This is the proposal's central vulnerability, and it is why the action-forcing mechanism is essential — without it, continuous auditing risks producing better-documented inaction.

A second objection concerns privacy and data security. A system that ingests government-wide payment and program data is a high-value target and a potential surveillance tool. Strong cybersecurity, data minimization, and limits on use are non-negotiable.

A third counterargument is the risk of false positives and audit overreach: aggressive anomaly detection can flag legitimate transactions, burdening programs and beneficiaries with unwarranted scrutiny. Human review and clear appeal processes are necessary to prevent auditing from becoming harassment.

A fourth consideration is that some apparent "duplication" reflects deliberate redundancy or locally tailored programs that serve real purposes; not every overlapping program is waste. Audits must distinguish genuine inefficiency from intentional design, which again requires professional judgment, not algorithms alone.

Finally, there is the modest cost and the meaningful institutional friction of standing up a new capability — though both are small relative to the spending at stake.

Conclusion

Public, continuous, AI-assisted auditing turns transparency into a fiscal instrument. The opportunity is well documented: $160 billion-plus in annual improper payments, hundreds of billions in estimated fraud, and GAO's own finding of $100 billion or more in achievable savings from duplication alone. Modern analytics make it possible, for the first time, to examine the full population of federal transactions continuously rather than sampling them once a year, and Treasury's AI payment-integrity results show the approach already works at scale.

The decisive caveat is that auditing identifies waste but does not cut it. The $100 billion target is realistic only if continuous auditing is paired with binding action-forcing mechanisms and genuine public transparency that together convert findings into recovered dollars. The audit infrastructure is cheap; the political will to act on what it finds is the real cost — and the real test. Built with independence, strong privacy safeguards, human judgment in the loop, and teeth behind its findings, continuous public auditing is among the highest-leverage transparency reforms available.

Sources

← Back to The Great Reinvention