AI Audit
Audit moves from sampling to full-population testing
Today: audit relies on sampling — examining a fraction of transactions and inferring the rest, because reviewing everything by hand is infeasible.
The next five years: full-population testing becomes standard. Every transaction can be screened, anomalies flagged, and the audit becomes continuous rather than annual.
The AI relation: models make 100% testing tractable; the auditor's value shifts to judgment on the exceptions the machine surfaces and on areas requiring professional skepticism.
Signal: sampling was a workaround for limited capacity. When capacity is unlimited, the entire audit methodology gets rewritten.