Claude for FP&A: Two Ways to Build a Faster Forecast
If you do any kind of financial planning, you know the two weeks I mean. Budget season lands, and suddenly you are rebuilding the same workbook you rebuilt last year, emailing department heads for numbers they have not thought about since March, and re-keying actuals into a template that breaks the moment somebody inserts a row. I have been teaching Claude for FP&A to finance teams all year, and the complaint is always the same: the analysis is not what eats the time. The assembly is.
The part that stings is that most of it is not judgment work. Pulling last year’s actuals apart to see what drove the number, typing three versions of the same forecast because the CFO wants a downside, writing the variance paragraph. That is mechanical work wearing a finance costume. Here are two things you can do this week — no new system, no consultant, no ticket to IT. Just your actuals export and about ten minutes.
Claude for FP&A, Tip One: Pull the Drivers Out of Your Actuals
Most budgets get built by taking last year’s number and adding a percentage. That is not forecasting, that is rounding. The better move is to figure out what the number is actually made of first, and that is exactly the kind of grinding Claude is good at.
Export your GL detail or P&L by month to CSV — twelve to twenty-four months if you have it. Attach it to Claude and type something close to this:
Here are 24 months of actuals by account and month. For each expense account over $50,000 annually, tell me: is this fixed, variable, or stepped? If variable, what does it appear to move with — headcount, revenue, or units? Show the month-over-month pattern that led you to that call, and list any accounts where the pattern is unclear.
What comes back is a driver table: account, behavior, likely driver, and the evidence. That last clause is the one that earns its keep — asking for the unclear accounts makes Claude flag its own weak spots instead of confidently guessing, and those flagged accounts usually turn out to have a reclass, a one-time charge, or a vendor change buried in them. You just got a review list for free.
Now you can build the budget from behavior instead of from last year plus four percent. Fixed accounts roll forward. Variable accounts get tied to the driver. Stepped accounts get a conversation about when the next step hits.
The gotcha: Claude only sees what you hand it, and a finance export is usually full of traps. If your CSV has subtotal rows mixed in with detail rows, Claude will read those subtotals as line items and your numbers will come back roughly double. Same for merged header rows and blank spacer rows. Before you upload, strip it to one flat header row and detail only. Thirty seconds of cleanup, and you avoid an answer that is wrong in a way that looks right.
Run Three Scenarios Without Building Three Workbooks
The second time sink is scenario work. Someone asks for base, upside and downside, and the usual answer is Save As, Save As, and now there are three files, three sets of assumptions, and no reliable way to explain why they differ.
Instead, describe the model once and let Claude do the arithmetic. Something like:
Base case: revenue $14.2M growing 6%, COGS at 41% of revenue, headcount 58 at average fully loaded cost $112,000, other opex $2.1M growing 3%. Build base, downside and upside. Downside: growth 0%, COGS 44%, no hiring. Upside: growth 12%, COGS 39%, 6 additional hires starting in Q3. Give me a table of revenue, gross margin, opex, EBITDA and EBITDA margin for all three, then a short paragraph on what has to be true for the downside to happen.
You get the table in seconds, and you get it with the assumptions written in plain English right next to the output. When the CFO asks in November why the downside assumed flat growth, the answer is in the conversation, not in someone’s head. Ask for it as a CSV or Excel file and it goes straight into your deck.
The gotcha: Claude will happily do the math, and you still have to tie it out. Before any of this reaches a board packet, pick one line — gross margin in the downside case is a good one — and recompute it by hand. If that ties, the structure is sound. If it does not, you found the bad assumption in a minute instead of in the meeting.
Plenty more on the calendar
If you want to see both of these run live, against a real export, with the messy parts included, I am teaching Advanced Claude for FP&A: Budgeting, Forecasting and Scenario Analysis on Tuesday, September 22. We go further than this post does — building the driver model, rolling it into a working forecast, and handling the variance write-up at month end.
If that date does not work, the full current schedule is on the front page of PCWebinars.com, with new dates going up as they are confirmed — Excel, Power BI, Copilot, ChatGPT and Claude sessions run most weeks. Register for this one, or browse the others.
About your trainer
I’m Tom Fragale. I have been a computer trainer and consultant for more than 30 years, and I have taught well over 30,000 people how to get more out of the software they already own — Excel, Access, Power BI, Word, Outlook, and now the AI tools that sit on top of all of it: Claude, ChatGPT, and Microsoft Copilot. My background is applied, not theoretical. Alongside teaching, I build real databases and reporting systems for clients in Excel, Access, and Power BI, so the examples I use in class are the problems I ran into last week, not textbook exercises. I also publish free tutorials on YouTube and longer courses online.
If any of this was useful, there is a lot more where it came from — come see what is coming up at PCWebinars.com.
