Six years ago I billed €150 an hour to do exactly what Claude does here in 20 minutes. If you're an auditor, an accountant, or you sign the audit fees at year-end, this is worth your attention, because the way audits get done is changing, and most firms are pretending it isn't.

The case: Baltic Components SIA, a fictional distributor of cables, connectors and sensors, ~€9M revenue, March 31 year-end, built with findings I've personally seen in client books. Three textbook procedures from the ISA rulebook (ISA 240, 330 and 520), a real general ledger, 76 PDF invoices, and two years of financial statements.

Procedure 1: Journal entry testing (ISA 240)

The fraud-risk classic: entries on weekends, round numbers, manual postings to revenue, entries by people who shouldn't have access. The standard approach is scrolling through 500 ledger rows in Excel.

I gave Claude the ledger plus an HR access list (including terminated employees) and the procedure description. Minutes later: a color-coded risk workbook, red/orange/yellow, with findings like a round €10,000 posted before business hours with no approval, an identical amount posted three times manually, and €15K to "other income" on a Saturday. Each flag explains why it's suspicious, so the auditor knows exactly what documents to request. Now imagine 10,000 entries instead of 500.

Procedure 2: Revenue cut-off (ISA 330)

The easiest place to dress up year-end numbers: ship in April, backdate the invoice to March. The procedure, tying every invoice around year-end to its actual delivery date, used to take a junior auditor two full days.

Claude cross-checked the sales register, the delivery notes and all 76 PDF invoices, and produced an issues table: March invoices for goods shipped and signed for in April, a brand-new customer with no purchase order and no delivery note (€24K of suspicious revenue), and goods shipped in March but invoiced in April. Net cut-off adjustment: roughly €83–95K of misstated revenue, without me opening a single invoice.

Procedure 3: Analytical review (ISA 520)

Compare this year to last, explain every ratio that moved more than 10%. Claude computed the ratios, flagged the movers, and, before I finished reading the prompt, surfaced the classic red-flag profile: revenue up 15% but receivables up 85% and inventory up 54%, working capital growing 3–5× faster than sales. It traced 37% of the entire revenue increase to one brand-new customer with a pending credit check and zero payments received, and told me exactly what to prioritize.

What changes (and what doesn't)

Someone still signs the file. Someone still sits across from the CFO and asks why €15,000 hit other income on a Saturday night. Claude doesn't do that conversation, the auditor owns the judgment, the risk and the signature.

What changes is the cost structure. A junior's entire week of work is now 90 seconds of prompting and a coffee. The firms that figure this out first will bill less, deliver faster, catch more, and outcompete the ones that don't.

The experiment: on your next engagement, pick one procedure. Run it the old way, run it through Claude, compare.