Establish reliability before relying on a report
Information produced by the entity, commonly called IPE, includes reports and data used to perform or test controls. A reviewer needs a basis for relying on that information: whether it covers the intended population, contains accurate fields and reflects the correct period. A report that looks complete can still omit a business unit, apply an unintended filter or contain changes made after extraction.
For external audits under PCAOB AS 1105.10, the auditor tests the accuracy and completeness of company-produced information, or tests controls over its accuracy and completeness, and evaluates whether it is sufficiently precise and detailed for the audit purpose. Report parameters, record counts and source comparisons may support these procedures; none is a universal validation method. Procedures for a company’s SOX program should reflect the information’s intended use and the applicable control requirements.
Separate completeness from accuracy
Completeness asks whether all relevant records are included. Accuracy asks whether the information represents those records correctly. For a financial report, comparing totals to a source balance may be helpful, but it may not address a missing field or an incorrect date. Teams should define the purpose of each validation step and retain evidence that shows the comparison actually performed.
Start by documenting the report name, source system, extraction period, filters and any subsequent transformations. Then identify which checks address the fields and population needed for the control. walktru supports connecting this evidence to defined testing criteria and retaining a reviewable record. The goal is reproducibility: another person should be able to understand which version was tested and why it was considered suitable.
Testing checklist
- Identify the report, source and reporting period.
- Review filters, parameters and transformations.
- Compare relevant counts, totals or fields to source information.
- Retain validation evidence with the control test.
Example: a revenue report used in a review control
Suppose a monthly revenue review uses a system-generated report. Before evaluating the manager’s review, the team considers whether the report includes the intended entities and month. Useful support may include report parameters, extraction documentation and a comparison to a relevant source total. If a spreadsheet was used to reshape the data, document that transformation as part of the information chain.
A difference should be investigated rather than concealed by a matching grand total elsewhere. For example, the period could be correct while one entity remains excluded. Define which checks are appropriate for your test and have the reviewer assess unresolved differences. This example describes a methodology consideration; it is not a claim that walktru has a universal connector or automatically validates every financial report.
From reliable IPE to defensible AI conclusions
Traditional IPE questions remain important when AI is involved, but they are not the whole evaluation. In his article, Dor Ad introduces IPE-A, Information Produced by the Entity’s Agent, to explore the evidence, testing coverage and accountability needed before relying on agent-produced conclusions. IPE-A is the article’s proposed concept, not a defined PCAOB auditing-standard term. That question is distinct from report validation: a reliable input does not on its own establish that an AI conclusion used the right procedure.
walktru’s deterministic AI and SOX TRAIL® help connect inputs, testing activity and results. Evaluate both the underlying information and the logic applied to it. Begin with a representative report and a defined validation approach, then assess the quality of the resulting record with your SOX and audit teams. Customer data is not used to train models.