Move sampled transaction testing from spreadsheets to linked evidence

Start with the selected sample

The imported row provides the transaction or folio context: identifiers, dates, amounts, tax status, account information, and other fields used by the audit team. The auditor chooses which columns identify each sample and can retain the source data for reference.

Evidence pages are then associated with the selected item. This prevents a broad document summary from being mistaken for transaction-level testing and makes it easier to see samples with no assigned support.

Use consistent classifications without losing nuance

GovAuditAI normalizes results into Supported, Potential Exception, Missing Documentation, Needs Additional Review, and Not Applicable. Those labels support sorting and follow-up, but the supporting facts, criteria, observation, and missing evidence remain visible separately.

An auditor can accept the result, override it with a reason, or mark the item as needing additional support. An AI failure does not overwrite completed review work on other samples.

• Supported when available evidence addresses the test
• Potential Exception when evidence indicates a possible variance
• Missing Documentation when required support is absent
• Needs Additional Review when ambiguity remains

Export the auditor-approved result

The final export should agree with the decisions visible in the review workspace. GovAuditAI includes the auditor action, final classification, notes, override reason, and decision timestamp so downstream workpapers reflect the reviewed result.

Document-request text can be generated from unresolved evidence gaps and edited before it is sent. This shortens follow-up preparation while preserving human control over communication.