Where AI Fits in the FP&A Workflow
Corporate FP&A is a deadline business. Every month you pull actuals, explain what moved, update the forecast, and package it into something leadership can act on. Then the cycle starts again. AI tools like ChatGPT, Claude, and Microsoft 365 Copilot will not run the business for you, but they can take the slow, repetitive parts of that cycle and shrink them from hours to minutes. This lesson maps exactly where AI helps and where it does not, so you spend your effort in the right places.
What You'll Learn
- The FP&A tasks where AI delivers the biggest time savings
- The difference between AI as a drafting partner and AI as a calculator
- Which tool to reach for: chat assistants versus Copilot inside Excel
- A simple rule for deciding what to delegate to AI and what to keep manual
The FP&A cycle, broken into AI-friendly tasks
Think of your month as a sequence of jobs. Some are pure judgment. Some are mechanical. AI is strongest on the mechanical-but-wordy jobs and the structured-thinking jobs.
High-value AI tasks in FP&A:
- Drafting variance commentary from a table of plan-vs-actual numbers
- Turning a messy budget request email into a structured assumption list
- Summarizing a 20-tab model into three talking points for your manager
- Writing the narrative slides for a board or management deck
- Brainstorming the drivers behind a revenue line you need to forecast
- Reformatting and cleaning exported data so it is ready to model
- Stress-testing your own assumptions by asking AI to argue the other side
Tasks to keep mostly manual:
- The actual arithmetic of your model (use Excel or Sheets formulas, not an AI's mental math)
- Final sign-off on numbers that go to leadership
- Anything touching material non-public information in an unapproved tool
The pattern is clear. AI drafts, structures, and explains. Your spreadsheet calculates. You decide.
Drafting partner versus calculator
The single most common mistake new AI users make in finance is asking a chat assistant to do math in its head. If you paste raw numbers and say "what is the variance," a language model may approximate rather than compute, and an approximation in a forecast is a liability.
There are two safe ways to get accurate numbers:
- Do the math in your spreadsheet, then ask AI to explain it. Calculate the variance in Excel, paste the resulting table into the chat, and ask for commentary. The numbers are yours; the words are AI's.
- Use a tool that runs real code. ChatGPT's data analysis feature writes and runs Python to compute results, and Microsoft 365 Copilot in Excel can apply Python-powered analysis and build pivot tables from a plain-language request. These execute actual calculations rather than guessing.
Keep this distinction in your head for the rest of the course. When accuracy matters, the math happens in a calculation engine, and AI handles language and structure around it.
Which tool for which job
You do not need every tool. Here is a practical split based on what FP&A teams actually have access to.
Chat assistants (ChatGPT, Claude): Best for drafting commentary, summarizing, brainstorming drivers, rewriting for an executive audience, and analyzing an exported file. Claude tends to handle long documents and large pasted tables well. ChatGPT's data analysis mode is strong when you want charts and computed results from an uploaded spreadsheet.
Microsoft 365 Copilot in Excel: Best when your data already lives in Excel and you want to stay there. It can summarize a sheet, suggest formulas, build pivot tables from a description, and outline a step-by-step plan before changing your workbook. If your company has Copilot licensed, this keeps sensitive data inside your Microsoft tenant, which matters for the data rules in the next lesson.
A reasonable starting setup: use Copilot in Excel for in-model work, and a chat assistant for the writing and thinking around the model.
A simple delegation rule
When you hit a task, ask yourself two questions:
- Is the output words or numbers? Words lean toward AI. Numbers stay in the spreadsheet.
- Would I be comfortable if a smart junior analyst drafted this for me to review? If yes, delegate it to AI and review the output. If the task needs your specific business judgment and cannot be checked quickly, keep it manual.
This rule keeps you fast without outsourcing the parts of the job that are actually yours. A forecast narrative drafted by AI and edited by you is a force multiplier. A forecast number invented by AI and shipped unchecked is a risk.
A quick example
Suppose you just finished your monthly actuals and have a plain table of revenue and opex by department, budget versus actual. A strong first move:
You are my FP&A analyst. Here is a plan-vs-actual table for May
(numbers already calculated in Excel). Write three bullet points
of variance commentary for my CFO. Flag the two largest drivers,
keep each bullet under 25 words, and use a neutral, factual tone.
[paste table]
You did the math. AI did the writing. You will refine this exact pattern throughout the course.
Key Takeaways
- AI is strongest on the wordy and structured-thinking parts of FP&A: commentary, summaries, drivers, and deck narratives.
- Never let a chat assistant do model arithmetic in its head. Calculate in your spreadsheet, or use a tool that runs real code.
- Use Microsoft 365 Copilot in Excel for in-model work and a chat assistant for writing and analysis around the model.
- Delegate when the output is words and you can review it quickly; keep judgment calls and final numbers manual.