Study proposes verification-cost metric to evaluate AI reliability in accounting
Researchers argue AI model evaluation for accounting must measure verification effort, not just correctness—identifying a new failure mode where auditors miss errors within budget constraints.