From Fraud Signals to Defensible Decisions

Executive Summary

Fraud is not only a financial-control issue. It can divert public funds, delay essential services, weaken confidence in government, and increase the workload of investigators and program staff. The policy challenge is equally important: agencies must strengthen prevention while preserving lawful access, procedural fairness, privacy, accessibility, and clear avenues for review.

The scale warrants executive attention. The U.S. Government Accountability Office (GAO) estimated direct annual federal fraud losses of $233 billion to $521 billion using fiscal year 2018–2022 data. GAO also reported approximately $186 billion in estimated improper payments for fiscal year 2025 across 64 programs at 15 agencies. These measures are not interchangeable. Improper payments may result from fraud, error, insufficient documentation, or noncompliance with payment requirements.

The strategic response is not simply to purchase another detection tool. Agencies need an enterprise capability that connects data, policy, analytics, investigation, evidence, and governance. Infosys Public Services recommends a federated model: programs retain decision authority while shared services provide reusable data integration, identity and entity resolution, analytics, case workflows, model governance, and outcome measurement.

Executive proposition: Design the fraud capability around defensible decisions. Every alert should connect a risk signal to the relevant policy, supporting evidence, authorized human action, and an auditable outcome.

Executive Decision Snapshot

The Scale of the Challenge

Public sources provide several different views of fraud and payment integrity:

• Estimated federal fraud loss: GAO estimated $233 billion to $521 billion in direct annual financial losses, based on fiscal year 2018–2022 data.
• Estimated improper payments: For fiscal year 2025, 15 agencies reported about $186 billion across 64 programs. GAO stated that these estimates do not represent the full government-wide extent.
• Overpayments: About $153 billion, or approximately 82% of the fiscal year 2025 improper-payment estimate, consisted of overpayments.
• Cumulative improper payments: GAO reported about $3 trillion in estimates since fiscal year 2003.
• Consumer-reported fraud: The Federal Trade Commission reported $12.5 billion in consumer losses in 2024.

Where Public-Sector Fraud Risk Concentrates

A Better Goal Than “More Alerts”

A modern fraud capability should improve decisions, not simply generate more alerts. High false-positive rates can divert investigators from material cases and create avoidable friction for legitimate applicants, recipients, vendors, and staff. Leaders should therefore define the decision points that matter most:

• Triage: Which referrals warrant review?
• Prioritization: Which cases should investigators address first?
• Intervention: Which action is legally authorized and proportionate?
• Disposition: What evidence supports the outcome?
• Learning: What should change in policy, controls, data, rules, or models?

This decision-first approach changes the technology question. Data sources, models, and automation earn a place only when they improve a defined decision and can be monitored for accuracy, fairness, timeliness, and operational value.

Target Capability: From Data to Evidence

1. Data integration and permitted sharing. Connect relevant program, payment, procurement, identity, vendor, and case data under explicit legal authority, purpose limitation, access controls, lineage, retention, and quality rules.

2. Identity and entity intelligence. Resolve people, businesses, addresses, devices, accounts, and relationships while maintaining confidence scores and reviewable source provenance.

3. Rules, analytics, and machine learning. Combine policy rules, anomaly detection, network analysis, and supervised models. Use the least complex method that meets the decision need, and monitor drift and performance.

4. Risk scoring and prioritization. Rank transactions or cases using transparent thresholds. Do not treat a score as proof of fraud or as an automatic adverse decision.

5. Case management and workflow. Route alerts, record investigator actions, capture evidence, support collaboration, and preserve disposition and appeal information.

6. Explainability and evidence package. Provide reason codes, contributing indicators, source references, policy links, and a record of human review suitable for audit and oversight.

7. Executive measurement. Report prevented or avoided loss using an approved methodology, together with false positives, case cycle time, recovery, service friction, model performance, and control compliance.

Operating Model: A Federated Fraud and Program Integrity Center

A centralized Center of Excellence can create consistency, but public programs retain distinct statutes, risks, evidence standards, and service obligations. A federated model balances enterprise reuse with program accountability.

• Enterprise governance: Set risk appetite, decision rights, escalation thresholds, data-sharing rules, records obligations, model governance, and oversight reporting.
• Program ownership: Keep policy interpretation, eligibility authority, case disposition, and appeal responsibility with the accountable program.
• Analytics factory: Maintain reusable typologies, rules, features, models, testing protocols, entity services, and documentation.
• Investigation operations: Standardize triage, routing, evidence capture, quality review, closure codes, referrals, and feedback loops.
• Responsible use: Require privacy, security, accessibility, fairness, human oversight, and impact assessments proportionate to the decision risk.
• Performance assurance: Establish baselines, independent validation, benefits methodology, control testing, and transparent reporting.

How to Begin Without Overreaching

1. Baseline one program. Document current losses or exposure, control coverage, false positives, case volume, cycle time, recovery, service friction, and data quality.
2. Prioritize one decision. Choose a material, measurable decision such as duplicate-payment review, vendor anomaly triage, or identity-risk escalation.
3. Design governance before automation. Confirm legal authority, data use, human decision rights, explainability, records, accessibility, security, and appeal requirements.
4. Pilot with a comparison method. Test against a defined baseline or control group where feasible. Measure both fraud outcomes and impact on legitimate users.
5. Validate independently. Review model performance, evidence quality, benefit calculations, bias and disparate impact, and operational adoption.
6. Scale reusable services. Expand only after the pilot demonstrates value, defensibility, and manageable service impact.

Measures That Executives Can Govern

Financial: Validated prevented or avoided loss; recoveries; cost per reviewed case; benefit methodology and confidence range

Decision quality: Confirmed-hit rate; false-positive and false-negative rates where measurable; overturned decisions; evidence completeness

Operations: Time to first action; case cycle time; backlog; investigator productivity; referral quality

Service impact: Added verification steps; processing delay; abandonment; complaints; appeals; accessibility issues

Governance: Data-quality exceptions; model drift; unauthorized access; control compliance; overdue reviews; remediation status

Path Forward

Government leaders should treat fraud and payment integrity as a mission capability, not an isolated audit activity. The immediate priority is to identify one material decision where fragmented data or delayed review creates measurable exposure. Agencies can then establish governance, test a proportionate intervention, validate the results, and scale reusable services across programs.

The measure of success is not the number of alerts or the sophistication of a model. It is whether the agency protects public resources, supports timely and equitable service, makes defensible decisions, and learns fast enough to respond as risks change.

About Author

Ajay Kumar

Ajay Kumar is a technology leader with over 24 years of experience driving large-scale digital transformation initiatives, from initial project implementation to final system integration and enterprise architecture. As principal Technology Architect at Infosys Public Services, he leads the Cloud and Infrastructure practice and is instrumental in helping clients modernize their technology landscapes.

Mr. Kumar is known for his strategic vision and collaborative leadership. Ajay excels at aligning complex business objectives with innovative technical solutions. He has a proven track record of spearheading programs that leverage agile methodologies, digital experience platforms, and cloud technologies to foster business innovation.

He is passionate about mentoring and guiding teams, actively sharing his insights on emerging technology trends to build a culture of continuous learning and excellence.