Notes on AI in finance, drawn from recorded runs on a full set of company books.
Plain explanations for controllers, accountants and firms: what worked, what failed, and the checks to keep.
Articles
Start with the article closest to the decision in front of you.
AI for accountants: what it can do on the close, and where it fails
A plain look at AI for accountants, using recorded runs on a fictional company's October close: what matched the answer key, what did not, and the checks to keep.
AI course for controllers: eight weeks, coursework and a capstone plan
What an AI course for controllers should contain: live sessions, graded coursework on a full set of books, controls, and a plan you can present to your CFO. How ours works and what it costs.
Agentic AI in finance: what an agent did on a month-end close
Agentic AI in finance explained through recorded runs: an agent that drafted accruals, declined a relayed approval and refused a hidden bank-change instruction, and what that does and does not prove.
How to test an AI control against a hidden instruction in a vendor invoice
A step-by-step way for finance teams to test an AI accounts-payable agent against prompt injection in a vendor invoice, with the results of one recorded test and what they do not prove.
An AI policy for an accounting firm: client data, the AICPA Code and IRC §7216
What to put in an AI policy for an accounting firm: data classes, approved tools, tax return information, review of AI output and testing, with the rules a firm should check against primary sources.
Choose one workflow on November 17, and hand in its redesign, controls included, on December 14.
Request a place in the November course: four Tuesdays from November 17 to December 8, 2026, each with one 60-minute live session, recorded and posted within 24 hours. Training a team? Say so on the form.