The Finding
In April 2026, GAO published Payment Integrity: Agencies' Estimated Improper Payments Increased to $186 Billion in Fiscal Year 2025 (opens in a new tab) (GAO-26-108694). Fifteen agencies reported estimated improper payments of about $186 billion across 64 programs, up about $24 billion from the prior year. Roughly $153 billion, or 82 percent, was overpayments.
HHS runs the three largest programs. In GAO's Table 1, Medicaid reported $37.4 billion (6.1 percent), Medicare Fee-for-Service $28.8 billion (6.6 percent), and Medicare Advantage (Part C) $23.7 billion (6.1 percent). That adds up to $89.9 billion, by Avalon's sum of the three lines. HHS told GAO that Medicaid rose $6.3 billion because of more errors in eligibility redeterminations and provider screening while it phased out COVID-19 enrollment flexibilities.
GAO also says the total understates the problem. It leaves out programs agencies consider highly susceptible, such as TANF at HHS, which HHS does not estimate because of statutory limits. The report's open items are matters for Congress. It makes no recommendation to HHS.
Why It Keeps Happening
Medicaid is the clearest case. HHS builds its Medicaid figure as a weighted average of states' estimates across three parts: fee-for-service, managed care, and beneficiary eligibility determinations. The figure rolls up state estimates, so the work behind it spans federal and state processes.
Documentation adds to the count. When a review cannot tell whether a payment was proper because records are missing or insufficient, the payment goes into the improper payment estimate anyway.
Oversight is uneven too. For fiscal year 2024, half of the 24 agencies that reported 99 percent of government-wide estimates fully complied with the Payment Integrity Information Act, according to their inspectors general. Inspectors general recommended fixes for inadequate risk assessments at five agencies and unreliable estimates at seven.
GAO's own fixes sit with Congress. In March 2022 it made 10 matters for congressional consideration, and nine remain open. One asked for a permanent data analytics center of excellence. Congress extended the Pandemic Response Accountability Committee's analytics center to 2034, which GAO calls neither permanent nor government-wide.
Why It Matters to Your Bid
Nothing in the report tells HHS to automate anything. The opening comes from the dollars and from the cause HHS gave. A $37.4 billion Medicaid line with a stated cause of eligibility and screening errors invites one question: which manual steps feed those errors?
That question lands in technical volumes as "understanding of the requirement" on operations and program integrity task orders. Evaluators look for three things in the answer: a clean split between rules-based steps (moving, matching, and reconciling records) and decisions that turn on judgment; a baseline stated before any gain is claimed; and a path for each automation through the security office.
The risky bid promises a lower improper payment rate. The report supplies no basis for that promise, and the agency owns the rate. A claim measured in hours returned, against a baseline both sides signed, holds up.
Technical Approach
The method below comes from Avalon's Process Automation service. It describes how the work is done, at the methodology level, inside the agency's own environment.
- Inventory the process area with the people who do the work. Interviews and observation with the staff handling the work, not only their supervisors, produce a candidate list with the systems and inputs each one touches.
- Score each candidate on seven criteria. The suitability matrix covers volume, frequency, rule clarity, input stability, exception rate, system accessibility, and data sensitivity. A step that turns on caseworker judgment scores low on rule clarity and is not built. The assessment says so and, where judgment dominates, routes it to a separate AI and machine learning assessment.
- Baseline hours before anyone builds. Annual volume, hands-on minutes per item, and exception rate for each candidate come from agency counts and observed timings, in a workbook the agency maintains. The baseline measures hours returned. It does not convert them to budget savings, and it does not predict a change in the improper payment rate, which the agency's own error data governs.
- Design API first, with screens as the fallback. For each system a candidate touches, the assessment records whether an API, export, or file interface exists. Where one does, the Process Definition and Solution Design Documents specify it. UI automation is documented as the fallback for systems with no other path.
- Pre-screen for the ISSO. Each recommended candidate gets a bot account matrix (every non-person entity, its privileges, and how credentials are vaulted and rotated), a logging specification, and inputs for the change board's security impact analysis, written against NIST SP 800-53 Rev. 5 controls AC-2, AC-6, IA-5, AU-6, and CM-3. Data flows support the agency's privacy impact assessment where PII use changes.
Avalon prepares these inputs and the agency approves them. The work is not an authorization, and it does not certify an automation as compliant. Sample data containing CUI stays inside the agency's boundary.
What "Fixed" Looks Like
GAO has not defined closure for automation, because it has not asked for any. Its closure test for the open matters is legislative: Congress acts, and GAO updates the status.
For the process-automation slice of an HHS task order, these are targets:
- Every candidate in the process area scored, with the evidence for each score written down.
- An hours baseline signed by the process owners before any build.
- A security pre-screen for each recommended automation that the ISSO accepts for review without rework.
- Hours returned reported against the baseline, kept apart from any improper payment measure.
Status Check
Nine of GAO's 10 March 2022 matters for congressional consideration were open as of GAO's April 2026 report. GAO-26-108694 makes no agency recommendation to HHS to track. HHS's inspector general separately reported that HHS could not publish corrective action plans or reduction targets for TANF, which has no estimate.
Teaming Fit
| Field | Detail |
|---|
| NAICS | 541511, 541512 |
| Avalon's sub role | Automation candidate assessment, hours baseline, and ISSO-ready security pre-screen for one process area under the prime's operations or program integrity task order |
| Small-business status | Small business; HUBZone (in progress) |
| Likely vehicles | GSA MAS (SIN 54151S IT Professional Services); GSA OASIS+ |
If HHS operations or program integrity work is in your capture plan, Avalon teams as the sub that scores the manual candidates, baselines the hours, and hands the ISSO a security pre-screen for each automation the agency decides to build.