AI workshops for finance teams
Hands-on AI workshops for finance teams, built on their own close, reconciliations and reporting. The team learns to do the work with agents and check it like any workpaper, so it has more time to be a thought partner to the business.
Where it starts
Finance teams are busy closing the books, with little time left to learn something new.
Most of the work repeats every month, lives in spreadsheets, and has to be right. What leaders of these teams tend to see:
- reporting that takes most of the month, with the last week spent turning numbers into a board-ready pack
- people who assume anything with math in it has to be done by hand
- prompts that take longer to write than the task takes to do
- sensitive data, such as payroll, banking and investor information, that has to stay out of the wrong tools
What teams like this work on
Finance teams already review each other’s work.
A preparer and a reviewer, a source behind every number, totals that tie: that’s how good work with agents gets checked too. The sessions use it. One agent does the work, a second checks it against the sources, and code does the math.
Each workshop runs on the team’s own work. From finance teams so far:
Balance sheet reconciliations prepared across a workbook, then checked by a separate reviewer against the bank and billing records, with anything unusual flagged before the review.
The general ledger checked for misclassifications and missed accruals against prior periods.
Quarterly investor reporting cut from about two hours to about 15 minutes, and a monthly management pack drafted with its charts and commentary.
Commission numbers pulled from the CRM and checked against last quarter’s known-good sheet.
A weekly list built from the aging report, billing and email, with a draft ready for each account owner.
What stays human
Finance teams keep the sign-off.
- Review of every number before it’s used.
- Judgment calls such as accruals and tax, and anything that needs a professional’s view.
- Sensitive data. Payroll and banking stay out, connections start read-only or draft-only, and nothing posts or sends without a person.
What changes
The work gets faster, and the checks get stronger.
At Flare, someone on the finance team rebuilt the quarterly investor reporting that used to take two hours, and it now takes about 15 minutes. He automated the monthly attorney review deck too, which had been hours of pulling data and formatting slides.
On another finance team, an accounting manager who had assumed anything with math in it had to be done by hand ran the annual book-to-tax reconciliation with AI, and it tied to the dollar.
Questions
Can AI do the math reliably?
Not on its own, so the sessions don’t ask it to. The agent writes code to do the calculations, works from the source files, and shows where every number came from.
How do we check output when we don’t already know the answer?
With the checks finance already uses: totals that tie to the ledger, a second reviewer that sees only the sources, a test run on a month that’s already closed, and two runs compared.
What happens to payroll and other sensitive data?
It stays out. Teams work from exports, or from connections set to read-only or draft-only, and check with their IT and legal leads before any real data goes in.
Can it run every month without someone starting it?
Yes. A task the team repeats becomes a saved skill, then a scheduled run, with the output waiting for review.
Other teams
Start with a call.
Thirty minutes about your team and its work. If a workshop would help, Harry would be glad to put one together.