Split Purchase Detection Root Cause Classifier
Description
Classify exceptions and audit findings associated with split purchase detection into root-cause categories such as data quality, interface timing, manual error, policy gap, or system limitation. The MVP would connect ERP transactions, vendor master, travel, payroll, contract pay, public/exclusion lists, investigative data and produce read-only recommendations for DoD OIG, DFAS, Components, DCAA/DCMA where applicable.
AI / analytics pattern
NLP classification + clustering
Automation level / stage
analytics triage
Expected benefit
Better remediation targeting, fewer recurring errors, clearer NFR/CAP analytics.
Audit / financial statement impact
Expenses, accounts payable, accounts receivable, compliance reporting
Controls / human review
Human approval required before posting, payment, denial, personnel action, or official audit response; model validation; drift monitoring; exception sampling; full prompt/data/output logging.
Data needed
ERP transactions, vendor master, travel, payroll, contract pay, public/exclusion lists, investigative data; master/reference data; audit logs; policy/control requirements; prior exceptions; relevant document evidence.
Possible metrics
root-cause coding accuracy; CAP targeting cycle time; recurring issue reduction
MVP scope
Start with one Component/reporting entity and one subprocess (split purchase detection) for two close/audit cycles; read-only outputs first.
Related material weakness / control objective
Improper payment prevention; fraud risk management