{"stable_id":"37f86dbf3a4048b3","slug":"us-treasury-check-fraud-ml","company":"U.S. Department of the Treasury, Bureau of the Fiscal Service","workflow_name":"Near-real-time machine-learning prioritization of potentially fraudulent Treasury checks","function_code":"financial_crime","pattern_codes":["threshold_as_control","exception_based_operations","queue_elimination"],"changed_assumption":"That potentially fraudulent checks must be found through slower post-payment reviews rather than prioritized in near real time.","evidence_strength":"reported","publication_tier":"showcase","freshness":"current","reviewed_at":"2026-08-23","updated_at":"2026-08-22","source_quality_summary":"2 primary; publication outcomes are reported.","caveat_summary":"Treasury-reported attribution. The agency does not publish model precision, false-positive rate, or a causal decomposition from other process changes.","freshness_caveat":null,"workflow_summary":{"before":[{"actor":"Payment-integrity analysts","action":"Review payment and bank information through existing fraud processes.","actor_type":"control"},{"actor":"Recovery teams","action":"Pursue suspicious payments after identification.","actor_type":"control"}],"hinge":"That potentially fraudulent checks must be found through slower post-payment reviews rather than prioritized.","after":[{"actor":"Machine-learning fraud process","action":"Scores and prioritizes potentially fraudulent Treasury checks in near real time.","actor_type":"ai"},{"actor":"Analysts and partner institutions","action":"Investigate prioritized cases and expedite recovery actions.","actor_type":"system"}],"decision_mode":"shared","decision_marker":"Human sets the threshold; AI decides each instance"},"before":[{"order":1,"actor":"Payment-integrity analysts","action":"Review payment and bank information through existing fraud processes.","handoff_to":"Financial institutions and law enforcement","control":"Rules, referrals, and manual prioritization"},{"order":2,"actor":"Recovery teams","action":"Pursue suspicious payments after identification.","handoff_to":"Financial institutions","control":"Recovery procedures"}],"after":[{"order":1,"actor":"Machine-learning fraud process","action":"Scores and prioritizes potentially fraudulent Treasury checks in near real time.","handoff_to":"Office of Payment Integrity analysts","control":"Risk-based screening"},{"order":2,"actor":"Analysts and partner institutions","action":"Investigate prioritized cases and expedite recovery actions.","handoff_to":"Recovery and law-enforcement channels","control":"Humans determine investigative and recovery action"}],"decision_rights":"The model prioritizes risk; authorized officials and financial institutions decide whether and how to investigate or recover funds.","exception_path":"Flagged checks proceed through human review and institutional recovery processes; the public source does not disclose thresholds or appeal procedures.","removed_work":["Undifferentiated review of lower-risk checks","Delay between suspicious-pattern emergence and recovery prioritization"],"outcomes":[{"metric":"Fraud and improper-payment recovery attributed to expedited ML check-fraud identification","baseline":"Before the enhanced AI process at the start of FY2023","result":"$1 billion recovered in FY2024; Treasury separately reported $375 million recovered in FY2023","period":"Fiscal years 2023 and 2024","scale":"U.S. Treasury checks processed through the Office of Payment Integrity","attribution_caveat":"Treasury-reported attribution. The agency does not publish model precision, false-positive rate, or a causal decomposition from other process changes.","evidence_label":"reported"}],"executive_lesson":"The model did not replace fraud investigators; it changed which transactions reached them first and compressed the time available for recovery.","anti_pattern":"Equating dollars recovered with model precision or allowing risk scores to trigger adverse action without human review.","questions_for_leaders":["Does the model only prioritize, or can it block payment?","What precision and appeal data should be public?"],"collections":["queue-eliminated","regulated-autonomy"],"bundle_version":"1.0.0","bundle_fingerprint":"sha256:c23c6cc2b88153d008ea8fda928f632ce0011fc2d4c5036672a16e5d895bab93","canonical_url":"https://brianletort.ai/transformations/us-treasury-check-fraud-ml"}