What you will actually do
Review AI-generated workpaper interpretations against real audit standards.
The system parses uploaded audit workpapers and makes decisions: which procedures address which assertions, what constitutes sufficient appropriate evidence, and where a potential misstatement exists. You will tell us when those decisions are audit-defensible and when they are not.
Define ground truth across account types and audit areas.
Revenue recognition, accounts receivable confirmations, inventory observations, debt covenants, management estimates, related-party disclosures — each behaves differently under AU-C standards. You will tell us what a correct result looks like for each, and provide real workpaper examples where possible.
Challenge the AI’s risk and exception logic.
When the system identifies a transaction as potentially misstated, is it? When it concludes that substantive procedures are sufficient without further testing, should it? You provide the professional judgment the engineering team cannot.
Shape the product roadmap.
You will identify gaps between what auditors actually need and what the system currently produces. Those gaps become the next build cycle.