AI course for controllers: eight weeks, coursework and a capstone plan

A controller asked to produce an AI plan has a different problem from a student learning prompts. The plan has to survive questions from a CFO, an audit committee and an external auditor. It has to say which steps the AI does, which a person approves, what the data rules are and how anyone would know if the tool started getting things wrong. A course for controllers should be judged on whether it helps you answer those questions with evidence.

This article sets out what to look for in an AI course for controllers, and then describes how AI for Accounting Professionals from Ledger & Agent is built against that list. The comparison is ours, so read the first half as a checklist you can apply to any course, including ours.

Five things to look for

1. Applied work on a full set of books. Slides about AI are cheap. A course should have you run tools on a complete ledger, bank statement set, invoice register and payroll file, so you meet the messy parts: items that do not tie, definitions that differ and data that is missing.

2. Failures shown openly. If every demonstration works, the course is a sales pitch. You should see the AI get things wrong and see how the error was found.

3. Feedback on your own work. Watching is not the same as doing. Ask who reads your submissions, how fast they are returned and how much of the feedback is automated. If an AI drafts feedback, you should be told, and a person should be accountable for what you receive.

4. Controls and governance alongside technique. A controller's job is the control environment. A course should cover testing an AI step, re-testing after a model change, setting measures with baselines before go-live, and how the AICPA Code, IRC §7216, the FTC Safeguards Rule, SOX and COSO apply. The NIST Cybersecurity Framework 2.0 and the NIST AI Risk Management Framework are voluntary frameworks, not rules, but they are what many security questionnaires ask about.

5. A deliverable you can use. The best test of a course is whether you finish with something you can put in front of your CFO.

How AI for Accounting Professionals is built

The course runs for eight weeks. Each week has one 60-minute live session, recorded and posted within 24 hours, and mandatory coursework submitted online: readings, experiments, the week's project, a discussion post and a short reflection. You post in a discussion forum shared with the whole cohort. You plan on about two to five hours a week outside the session, with the heaviest weeks being 3 to 7.

The weeks move from the foundations to the close and then to risk: the jagged frontier of what AI does well and badly and what it costs, prompts and assistants, agents, the month-end close, reporting and audit support, security and privacy, governance and KPIs, and then ethics, people and the capstone.

The books. Everything runs on Northwind Supply Co., a fictional industrial-supplies distributor in Cleveland, Ohio, with about 95 staff and roughly $45 million a year in sales, under US GAAP. The October 2026 close runs through Weeks 3 to 7. The data is invented, so you can paste it into any AI tool.

The recordings. Every week includes a recorded AI run, eight in all, checked against an answer key. In Week 3 an agent prepared five accruals that matched the key and balanced at 169,373.31, then declined an approval relayed by another agent. In Week 6 an agent with write access to a vendor master refused a bank-detail change hidden in an invoice, and the vendor master was unchanged afterward. In Week 8 an AI reviewer flagged 13 of 14 planted errors. Each is one run on one model, and the course says so.

The marking. An AI grading assistant drafts feedback against a grading guide for each week. A person reviews every piece returned for revision, a sample of those marked complete, and every capstone score. You are told this, and you can ask for a person to look at any result.

The finish. You complete the course by doing the work: having all eight weekly submissions accepted and a capstone plan scored 70 or more out of 100. The capstone is built a section a week, so it is ready when the course ends, and it is written for your own finance function: the workflow you chose, the controls, the data rules, the measures and the rollout.

Where to start, and what it costs

Ledger & Agent's first course is AI You Can Sign Off On, a four-week course. The next cohort opens soon. Each week brings one 60-minute live session, recorded and posted within 24 hours, about one to two hours of coursework, and written feedback that an AI drafts, with a person approving every return for revision, a sample of completes and every capstone score. At the end of week four you hand in one workflow redesigned with AI, with its checks and controls, run on the Northwind books. It is the quickest way to find out what AI does to one piece of a close before you commit to the full plan.

The full program is $2,950 per seat. A firm or finance function that wants several people trained, or a cohort of its own, can contact us about team training. Refunds are in full until 14 days before the first session, and we publish that date on the enrollment page when the cohort's dates are set; after that your seat can move to a later cohort or pass to a colleague. A security and privacy pack is available on request for the person who approves the purchase.

Enroll now

---

General education, not legal, tax, audit or security advice. Confirm the rules (AICPA Code, IRC §7216, FTC Safeguards Rule, NIST frameworks, auditing standards) against current primary sources before you rely on them. Northwind Supply Co. is a fictional company and all of its data is invented for teaching.