Large-Sample Testing Evidence Support Assistant
Description
For auditor-selected samples, retrieve support, extract required attributes, complete tie-out tables, and route packages for reviewer approval.
AI / analytics pattern
document AI + RAG + workflow automation
Automation level / stage
assistive / analyst-in-the-loop / auditor-ready evidence support
Expected benefit
Reduces cycle time and rework for manual large-sample testing.
Audit / financial statement impact
Supports GAO-described manual testing of large samples with AI tools as needed.
Controls / human review
Human review required for AI recommendations; cite source records/documents; retain prompt/output logs, model/version metadata, reviewer approval, and exception rationale; monitor accuracy and bias/adverse impact where applicable.
Data needed
Advana-FM governed data products, source system extracts, transaction populations, GL/subledger detail, supporting documents, reviewer actions, data-quality rules, lineage metadata
Possible metrics
coverage %, completeness %, source-to-statement traceability %, sample package cycle time, first-pass acceptance %, exception aging, reviewer override rate, evidence gap closure rate
MVP scope
Build read-only MVP for one DWCF reporting entity / one line-item action team; validate data completeness, source links, exception queues, and human-review controls before production workflow integration.
Related material weakness / control objective
Material line-item support; source-to-statement traceability; data completeness; evidence reliability; large-sample testing support.