Free Resource
When a human did the books, quality control was invisible because it was baked into the doing. Once AI produces the books, the data still arrives — but nobody vouched for it. This kit gives you the first concrete artifact for making your checking visible: a correction log, an exception register, and a guide to running them.
The Problem
When a human reconciled the account, they stood behind it without thinking about it — quality control was free, baked into the doing. Take the human out of production and the data still arrives, but nobody vouched for it. A loan draw miscoded as revenue, a duplicated bill, an owner's distribution booked as an expense — these used to get caught on the way through. Now they can sail straight into the dashboard your client decides from.
The checking that used to be invisible has to become a deliberate job. And the value isn't the quality control you claim — it's the quality control you can show. This kit is the first artifact that makes yours visible.
What's Inside
Everything you need to start a correction record on your next review — and turn it into a moat your competitors can't copy.
Eleven columns. One row per catch — what the AI produced, what was wrong, the client-specific rule it missed, your correction, and the severity. Two worked examples included so you can see the level of detail that makes an entry useful.
Eight columns. When the same catch repeats for a client, promote it to a standing rule here — the context you feed back into the AI so it stops making that mistake. This is the layer that compounds.
A capture prompt you paste into Claude or ChatGPT, a worked example of feeding a correction back into AI context end-to-end, and a 90-day cadence to make the habit stick.
How It Works
The single biggest failure mode is "I'll log it later" — you won't. The capture prompt in the How-To turns a plain-English description of the catch into a structured Correction Log row in 30 seconds.
A one-off is noise; a repeat is a missing rule. The register is where the firm's knowledge actually accumulates — and it's where the compounding starts.
The whole moat lives here. A captured rule that never makes it back into the AI's context is a rule you'll re-catch next month. The How-To walks one real example end-to-end, from the catch to the encoded context block.
Sort the register by Times Seen, encode the top recurring rules, and tell the story when a client asks how you check the AI.
Stay Current
Subscribe to The AI Accountant newsletter and get the QC Starter Kit delivered to your inbox, along with weekly analysis of the AI developments that matter for your practice.
Your Move
The log is yours from day one — no one's permission needed. The compounding starts the day you make the first entry. Download the kit, open the spreadsheet, and write the first row this week.