AUDITOME // ASE
JUDGMENT TRAINING

Exercises

Real workflows. For each one, decide whether it has governance or only visibility — then reveal the diagnosis across the seven dimensions. This trains the judgment to spot the gap on the spot.

DRILL

The Autonomous Refund Agent

A support team deploys an AI agent that can issue customer refunds up to $500 automatically. Every refund is written to an activity log with a timestamp and the customer ID. A weekly dashboard shows total refunds issued. There is no step that checks the agent's authority limit at the moment of action, and refunds cannot be paused mid-run.

DECIDE FIRST

Does this workflow have trace, policy, authority, evidence, review, redress, and suspension?

DRILL

AI-Drafted Contract Clauses

An AI tool drafts contract clauses. A lawyer must click 'Approve' before any clause is sent to a client. The system records who clicked approve and when. However, it does not check whether that lawyer currently holds signing authority for that contract type, and approvals from former employees in the historical record were never invalidated.

DECIDE FIRST

Does this workflow have trace, policy, authority, evidence, review, redress, and suspension?

DRILL

The Pre-Trade Check

A trading workflow uses an AI assistant to propose trades. Before any trade executes, an engine evaluates the proposal against position limits and mandate rules, verifies the desk's authority, and emits a signed receipt for every outcome — approved, rejected, or escalated to a risk officer whose authority is checked. Disputed trades can be reconstructed from the receipt. The team is still building the ability to pause a strategy mid-execution.

DECIDE FIRST

Does this workflow have trace, policy, authority, evidence, review, redress, and suspension?

DRILL

AI Resume Screening

An AI screens job applicants and auto-rejects those below a score. Rejections are logged and a monthly report shows pass/reject counts. Rejected candidates have no way to challenge the decision, and the company cannot reconstruct why a specific candidate was rejected beyond 'the model scored them low.'

DECIDE FIRST

Does this workflow have trace, policy, authority, evidence, review, redress, and suspension?