U.S. government agencies estimate roughly $160 billion in improper payments in FY2024, bringing the cumulative total since FY2003 to approximately $2.7 trillion. The problem is structural, not behavioral, and the tools to address it look fundamentally different from what most agencies are using today.
The Numbers
In FY2023, the GAO's work resulted in $70.4 billion in financial benefits for the federal government. That is a return of approximately $84 for every dollar in their budget. Efficiency and fiscal stewardship are not abstractions. They are measurable results.
And yet, improper payments continue to climb. The federal government's own estimates put FY2024 improper payments at roughly $160 billion. Since systematic tracking began in FY2003, the cumulative total has reached approximately $2.7 trillion.
These numbers are not primarily about fraud. The vast majority of improper payments stem from three structural problems that no amount of guidance documents or training programs will resolve.
Three Structural Drivers
The Velocity Gap. The volume and speed of government outlays have expanded dramatically. When billions move under tight period-of-performance windows, traditional manual review becomes the primary bottleneck. Agencies are not slow because they are careless. They are slow because the volume of verification work exceeds human processing capacity.
Consider a mid-sized state agency processing 2,000 invoices per month against 300 active contracts. Each invoice requires rate verification against the current amendment schedule, scope confirmation, period-of-performance validation, and documentation completeness review. At 30 minutes per invoice, that is 1,000 person-hours per month of pure verification work. Most agencies allocate a fraction of that.
The System Silo. Legacy ERPs, grant portals, contract repositories, and accounting platforms do not communicate with each other. Data is re-entered manually. Reconciliations happen offline. The contract lives in one system, the invoice in another, and the payment authorization in a third. Verifying that all three are consistent requires a human to open multiple applications and manually compare records.
This is not a technology limitation in the traditional sense. Modern APIs and integration patterns can solve it. But the procurement for a new integrated system takes 18-36 months, and the legacy systems have 10-20 years of institutional knowledge encoded in their workflows.
The Verification Burden. For a payment to be compliant, it requires knowledge of base contracts, amendments and change orders, grant terms, multi-year appropriations language, and evolving eligibility rules. At scale, this level of forensic review is nearly impossible to maintain without error.
The practical result: agencies verify what they can in the time they have, prioritize large-dollar transactions, and accept risk on the rest. This is rational behavior given the constraints. But it produces the $160 billion figure.
Why the Strategy for 2026 Cannot Be "Hire More People"
The arithmetic does not work. To manually verify every invoice at the level of rigor that audit standards require, most agencies would need to triple their contract management staff. Government hiring timelines, compensation constraints, and competition with the private sector for qualified professionals make this unrealistic.
The alternative is to change the nature of the verification work itself. Approximately 70% of the checks required for invoice compliance are deterministic: does this rate match the current contract schedule, is this date within the period of performance, does this line item appear in the authorized scope? These checks do not require judgment. They require access to the right data and the ability to compare it consistently.
When the deterministic checks are automated, the remaining 30% that require human judgment can receive the attention they deserve. Officers spend their time on genuinely ambiguous cases, not on rate lookups that a system could perform in milliseconds.
State-Level Implications
The federal numbers capture attention, but the dynamics are identical at the state level. Texas processed over $89 billion through HHSC alone in FY2024. Florida's AHCA administered $33 billion in Medicaid payments. Virginia's state agencies collectively process billions in vendor payments annually.
Each of these jurisdictions has audit findings that trace directly to the three structural drivers. The SAO in Texas, the Auditor General in Florida, and the APA in Virginia all document the same categories of failure: rate discrepancies, scope authorization gaps, and documentation deficiencies.
The pattern is not a coincidence. It is the predictable result of asking humans to perform verification work at a scale that exceeds human processing capacity. The question is not whether to automate pre-payment verification, but how quickly agencies can deploy it.