A finance leader I work with had never once opened an AI tool. Not ChatGPT, not Claude, nothing. Well, that may not be completely true, they probably used the free version of ChatGPT, or maybe even had used their company's AI tool but really only for Google-like search queries, answer this question. Last week we sat down with their monthly board pack, forty-odd pages of it, and in under an hour turned it into a single page the board could actually read. This is the post I would hand every finance leader who tells me they are curious about AI but have no idea where to start. If that is you, AI for board reporting is the right first task, and here is exactly how to do it.
I am not writing this for engineers. I am writing it for the CFO, the controller, the finance director who has heard for months that they should be using AI before they are replaced, feels a bit behind, and does not want to be talked down to. You are not behind. You are about to do your first useful thing with AI, and it happens to be one of the most painful jobs in your month.
Why board reporting is the right first task
Three reasons I always start finance people here.
It hurts every month. Pulling a readable summary out of a forty-page pack is hours of work, and it is the same hours every single month. Anything that repeats is worth automating first.
It is low risk if you do one thing right, more on that below. You are summarising numbers you already have, not letting a machine touch your ledger.
And the win is visible immediately. You hand your board one clean page instead of forty, and they notice the same day. A first win you can see is what makes the second task happen.
The three steps
Here is the whole thing. No jargon.
Step one. Open a tool. I would start with Claude or ChatGPT on the paid tier, because the free tiers cap how much you can paste in. Start a new chat. That is the entire setup.
Step two, and this is the one that matters. Take out anything you would not want a stranger to read before you paste it in. Salaries by name, anything personal, anything under NDA. Either delete those rows, or use a tool with a business data setting that keeps your input out of training. I treat client data the same way, and I wrote about how I keep a client's information out of a model if you want the careful version. The grandmother test: if you would not read it aloud to a stranger on a train, take it out first.
Step three. Paste the pack and use this prompt. Copy it word for word.
"You are my finance assistant. I am going to paste my monthly board pack. Give me a one-page summary with four parts: the headline numbers versus last month and versus budget, the three things going well, the three risks, and the three decisions the board needs to make. Plain English, no jargon, short sentences. If a number looks off or is missing, tell me instead of guessing."
That last line is the important one. It tells the tool to flag gaps rather than invent a tidy answer, which is the single biggest worry finance people have, and rightly so.
Then one follow-up to tighten it.
"Now cut it to half the length. Keep the numbers and the decisions. Lose the adjectives."
That is it. Forty pages to one, in three steps.
What good looks like, and what to watch
The first pass will be about eighty percent right. Your job is the other twenty. Read it against the pack. Where it is vague, ask it to be specific. Where a number is wrong, tell it the right one and ask it to redo that section. You are the controller here, not the passenger. The tool drafts, you check, the same way a junior analyst hands you a draft you would never send untouched.
Do not skip the check because the writing reads well. AI writes a confident wrong number in exactly the same tone as a confident right one. That is exactly why a finance person, not the machine, signs off.
If you would rather not do it yourself
If you are a CFO or finance director reading this and thinking it sounds useful but you do not have an afternoon to learn it, that is most of who I work with. I build this for finance teams: the prompts, the safe setup, the repeatable monthly version, so the one-page summary writes itself every month and your team checks it instead of building it. It is the same work behind the AI financial controller I have been building in public, the one that caught half a bank statement a human had missed. Board reporting is just the easiest place to start.
A few quick questions I always get
Is it safe to put financial data into AI? Yes, if you redact what is sensitive and use a business tier with training turned off. The risk is not the technology, it is pasting things you should not. Decide that up front.
Which tool should a finance team start with? Claude or ChatGPT on a paid plan. Pick one, learn it well, ignore the rest for now.
Will this replace my finance team? No. It replaces the boring half of their month so they spend the rest on the work only they can do. The person still signs off.
How long until I trust it? A few cycles. Run it alongside your normal process for two or three months, compare, and let trust build on evidence, not hope.
I am putting the exact prompts and a short before-you-start checklist into a free starter pack for finance leaders. It will be on the blog and on Substack. Subscribe and I will send it over.
Handling other people's numbers carefully is not abstract for me. It is the same care I owe every person who gives to the fund I run in my son's name. Your board deserves the clean page. Your data deserves the careful setup. You can have both, and you can start this afternoon.
Monthly Revenues $11,000 | Clients 2 | Prospects (AI outbound agent now live) | Team: Me + part time Jan (CTO) Day 82 of 365.