Method
How adubio produces and rates findings.
Severity rubric
Severity answers one question: if this finding is valid, can the workpaper's stated conclusion still be relied on as written?
Four checkable flags decide the rating
Conclusion-linked
Does the finding directly affect the workpaper's stated conclusion? Yes or no.
Evidentiary status
Is the evidence contradicted, unsupported, or supported with imprecision?
Citation materiality
Is the substance of the citation wrong, its applicability unclear, or is the issue one of precision only?
Internal consistency
Does the workpaper contradict itself? Yes or no.
The conclusion cannot be trusted until the finding is resolved.
The conclusion is weaker than it should be but not shown to be wrong.
Worth fixing without changing what the workpaper can be relied on for.
Every rating carries a one-line rationale stating which flags were set. The auditor can accept, amend or reject any finding.
Synthetic institutions
Qlogue builds fictional but coherent banks, with known defects, to test AI-assisted assurance under reproducible conditions.
These synthetic institutions — including Meridian, the bank used to validate adubio Analyse — contain interconnected policies, controls, workpapers and known defects. Testing against a controlled environment with a known ground truth allows Qlogue to evaluate whether adubio's findings are accurate, consistent and appropriately rated across repeated runs.