What “governed” means here
Governance is not a dashboard logo. It is a set of decisions recorded before, during, and after a model touches ledger data: which accounts are in scope, who may change thresholds, how exceptions are cleared, and how conclusions are worded so a reviewer can reconstruct judgment without replaying the software.
Scoping before screening
Teams that skip scoping invite the model into judgments it cannot own — going-concern assessments, related-party substance, or fraud brainstorming conclusions. We teach a short scope memo that lists permitted screening zones and explicit human-only zones, then ties both to the audit plan’s risk assessment.
Exception control
Raw hit lists are not evidence. A governed queue ranks exceptions by engagement risk, samples cleared items, and requires a one-line rationale for overrides. Seasonality and known system cutovers are documented as expected patterns, not as silent suppressions.
Evidence standards for Korea-based files
Working papers serving Korean group calendars often need bilingual labels and denser cross-references. Our guidance stresses version IDs, threshold tables, and reviewer identity so EQCR and network reviewers can follow the trail when tools update mid-season.
Where training fits
Policy alone rarely changes Friday-night habits. Courses such as Governed Ledger Review put teams through planted ledgers so they practice saying no to confident but wrong model output.