Selected claims in an assurance report
The current case defines two claims for review under one exact source and rule context.
Capability 03 / AI Review & Exception Queue
Auxtho records how claims entered review, compares each one with exact source proof, preserves the machine result and reason, and routes the items that need judgment to the accountable reviewer.
The reviewer receives the disputed claim, exact source correction, and reason together, then accepts, corrects, holds, or rejects it.
The job in plain language
Auxtho performs the repeatable comparison and prepares the exception. The person decides how the disputed claim should change.
The current case defines two claims for review under one exact source and rule context.
Auxtho binds each finding to the exact artifact, source proof, reason code, and queue order.
The reviewer resolves the prepared exception without erasing the original machine result.
The queue shows why the claim is overbroad and the exact source wording that restores the supported statement.
The record preserves both stages and the exact input bindings that must remain current before authorization.
Actual product screens
The first panel shows the exception after the accountable reviewer applies the exact correction from the source. The second shows the resulting review checkpoint.
Review flow
The run identifies whether each claim was user-specified, workflow-defined, rule-selected, or model-decomposed.
The machine result retains source identity, page, wording, passage hash, and the reason behind its classification.
Deterministic ordering sends supported work forward and presents exceptions with their reasons and owner role.
The reviewer accepts, corrects, holds, or rejects the prepared result; the original finding remains preserved.
Changed artifact, source pack, rule, claim set, route, target, or authority invalidates the previous review binding.
M122 AI_REVIEW_EXCEPTION_QUEUE_GO in terminal tree 1cca2c79d92ffba686883dd0ef3a8c4eb08f7c1b. Current revalidation included the 115 M121-M125 focused backend tests, TypeScript, focused M122 lint, and the real local App/API/transactional-store/human-correction browser path. Proof boundary: a local provider-free two-claim synthetic corpus; it does not establish automatic identification of all material claims, a workload automation percentage, provider-backed execution, customer effectiveness, production use, or compliance certification.