The question used to be whether finance teams should bother with AI.
That question is over.
According to Gartner’s 2025 AI in Finance Survey, 59 percent of finance leaders now report using AI in their function. That figure has more than doubled in two years. Which means if you’re not using it yet, more than half your peers already are.
The real question now is how to use AI in finance well.
Not just how to type a prompt into ChatGPT. That bit is easy. I mean, how to use it in a way that genuinely makes you a better finance professional, that won’t damage your credibility, and that doesn’t let the business work around finance entirely.
Because here’s the part most of the AI articles skip over.
Using AI badly in finance is worse than not using it at all.
A spreadsheet with an error is embarrassing. A board paper with a fabricated variance driver is career-altering. And in a function that’s built on trust in the numbers, you only get to make that mistake once or twice before people stop bringing the interesting work to you.
So this guide is for both audiences I work with every week. The individual FP&A analyst, finance business partner, or finance manager who wants to start using AI tomorrow morning. And the CFO or head of finance trying to work out how to roll it across a team without something going wrong.
At the FBP Team, we say — “once the robots arrive, the only thing left will be business partnering.”
This guide is about getting the robots in well, so the partnering bit you do afterwards is the strongest version of itself.
What AI Can Do in a Finance Function (And What It Can’t)
Before you use AI, you need an honest picture of what it’s good at.
The thing to remember is that tools like ChatGPT, Copilot, Claude and Gemini are linguistic, not analytical. They are very good at predicting what words probably come next based on patterns they’ve seen in training data. They are not good at understanding what EBIT means, or why a working capital movement matters more than a profit one.
That distinction matters more than people realise. It explains both what AI does brilliantly and where it falls over.
What it does well in finance work:
- It is fast at the grunt tasks:
- Cleaning a messy dataset
- Writing an Excel formula
- Drafting an email
- Summarising a 30-page document
- Turning bullet points into a paragraph
- It is good at frameworks: Give it a half-formed analysis, and it will structure your thinking.
- It is good as a thought partner: Ask it to argue the opposite view of your recommendation, and it will give you something you can pressure-test.
What it struggles with:
- It cannot see the context you haven’t given it.
- It does not know your business model, your customers, your cost structure, covenants, or the conversation you had with sales last week.
- It does not know what is missing from your dataset.
- It will fill gaps confidently with plausible-sounding nonsense (called AI hallucinations), and it will not flag that it has done so.
Which means using AI well in finance comes down to two skills the smartest professionals are already developing: knowing what to feed it, and knowing how to interrogate what comes back.
Gartner’s research is striking on this. Their 2025 priorities survey found that only 36 percent of CFOs feel confident in their ability to drive AI impact, and a separate report found that 86 percent of finance teams said they had achieved no significant value from their AI investments. That is not because the technology doesn’t work. It is because most teams have not learned how to use it in a way that maps to actual finance work.
Which brings us to the practical stuff.
Where to Start as an Individual Finance Professional
If you are an analyst, a business partner, a controller, or a finance manager, this is the section for you. The bar to start using AI is low. You don’t need a transformation project, a steering committee, or a vendor contract. You need a free or paid account on ChatGPT, Copilot, Claude, or Gemini, and about an hour to experiment.
The trick is to start with use cases where the cost of being wrong is low and the upside in time saved is high. Here are the six I see working best in practice.
1. Cleaning messy datasets
You know the file. Three different formats for dates, names spelled four ways, currency symbols pasted into number columns, blank rows scattered through. Paste a sample into AI and ask it to write the Power Query or Excel logic to clean it. It will take you ten minutes to do what used to take half a day. The output is auditable because you can see the formula.
2. Writing Excel, Power Query, and SQL formulas
This is the single highest-return use case for most finance teams. Instead of waiting four sprints for IT to pull a query, write a plain-English description of what you want, paste your column names, and ask AI to draft the SQL, DAX, or Power Query M. Run it against a small sample first to check. The time saved compounds. Finance professionals who use AI this way report a meaningful uplift in throughput.
3. Drafting financial commentary
This one needs care, which is why I’m flagging the caveat now rather than later. AI is excellent at turning raw variances into a readable narrative. It is terrible at knowing which drivers are real. Use it to draft commentary based on data you supply, never to generate commentary it has invented. The discipline is to write the analytical insight yourself in dot-point form, then ask AI to turn your points into board-ready prose. Not the other way around.
4. Stakeholder emails and meeting prep
The email to the operations director explaining why their cost centre is over budget. The note to the CEO summarising what came out of the forecast meeting. The agenda for next week’s commercial review. AI drafts all of these in under a minute, and you spend the saved time on the conversation itself rather than the wording. This is where most finance professionals see their first real productivity gain.
5. Pricing and scenario modelling support
For ad-hoc questions like “if we drop price by eight percent and volume grows by twelve, what happens to gross margin?” AI will set up the calculation logic, build the sensitivity table, and explain the result. Always check the maths against your own model. But for the back-of-the-envelope work that used to consume an afternoon, it is a different speed of working.
6. Document and meeting summarisation
Paste in the 40-page board pack, the audit report, the transcript of the quarterly review, and ask for the five things finance should care about. It will surface what matters. This is one of the use cases most adopted across finance functions in 2026, alongside knowledge management and accounts payable automation.
If you are good with prompts, you can even build your own custom GPT with the right kinds of questions to ask the data and the business. It is a useful starting point for anyone learning what good prompts in this domain look like.
A final note for individuals. Build the habit of treating every AI output the way you would treat work from a clever new graduate. Useful, fast, often correct, occasionally confidently wrong about something important. You sign your name to what goes out. The AI does not.
How to Roll AI Out Across a Finance Team
If you’re a CFO, head of FP&A, or finance director, the calculus is different. You’re not just asking how do I use this myself, you’re asking how do I get my team using it well, consistently, and without something going wrong on my watch.
The honest answer is that most finance functions are not getting this part right yet. Go back to Gartner’s research, and you’ll see that 77 percent of CFOs cite a lack of technical skills within finance as the critical reason their function has not yet adopted AI, and 86 percent of those who have adopted it report no significant value.
In the teams I work with, three reasons keep coming up.
Address the data before rolling out the tools
If your master data is wrong, AI is useless. I have a full article explaining why data quality in AI causes finance projects to fail, which I’d recommend reading if you’re planning a rollout. The summary version is this. AI sitting on top of bad master data does not produce better insights. It produces faster, more confident, harder-to-spot-wrong insights. The work to clean customer hierarchies, cost centre mappings, product groupings, and account structures is unglamorous and unavoidable.
The honest test is this. If you would not run a manual forecast off your current master data without a sense of dread, you should not let AI automate it.
Focus on the skills needed to use the tools
Buying licenses for Copilot or rolling out Microsoft Fabric does not give you an AI-enabled finance function. It gives you finance professionals with new buttons to press. The actual capability comes from people knowing which questions to ask AI, how to structure a prompt, how to validate output, and, crucially, how to spot when something has gone wrong. Skills before tools, every time.
The team treats AI as a productivity play, not a role-redesign one.
This is the strategic miss. If AI takes 30 percent of the grunt work off a finance team, that 30 percent has to go somewhere. The teams that win are the ones that redirect it into commercial conversations with the business to help your organisation make better decisions. The teams that lose are the ones that absorb the saved time, deliver the same outputs faster, and then get downsized in the next restructure because nobody can articulate what the team is doing with the extra capacity.
This last point is the one that keeps me up at night for the profession. It’s the subject of our piece on finance’s AI wake-up call, which explains why the risk of AI to finance careers isn’t replacement — it’s being bypassed entirely.
If you’re planning a rollout, the practical sequence I’d recommend is:
- Start with a small group of curious volunteers across grades, not just senior staff.
- Pick three to five high-volume, low-risk use cases, such as commentary drafting, dataset cleaning, or document summarisation.
- Set explicit ground rules about what AI is not allowed to do without human review (anything that goes to the board, anything that involves a financial decision, anything that involves customer or employee data).
- Run for 60 days.
- Measure time saved per person per week.
- Then decide what to scale.
That sequence sounds slow. It is. It’s slow because the cost of an early high-profile error (like an AI-generated number in a board paper that turns out to be invented) sets the function back two years. You can’t afford that.
The Risk of AI Getting Things Confidently Wrong
AI models will generate output that sounds plausible. Fluent, well-structured financial-sounding output that isn’t grounded in the data you gave it. It will attribute variances to drivers that do not exist in your file. It will reword “cash outflows increased due to project investment” as “cash outflows increased due to higher capital expenditure” and not flag that the meaning has subtly changed.
In most industries, that’s a footnote. In finance, it costs you credibility.
The protective habit is to treat AI as a draft, not a final answer, and to understand the level beneath every number it produces. You should be able to substantiate every number and every claim AI produces for you. If you can’t, don’t send it.
Will AI Replace Your Finance Job?
The short answer is no. The longer answer is more uncomfortable.
AI will not replace finance professionals who can translate numbers into commercial decisions. It will replace (or more accurately, route around) finance professionals whose role is to produce reports and hand them over.
The difference between those two finance people is not technical. It’s not about who knows more about AI tools. It’s about who can sit in a room with a sales director, an operations leader, and a CEO and make the financial implications of a decision land in a way that changes what they do next.
The real career risk isn’t being replaced by AI but being bypassed by the business because they’d rather ask ChatGPT than book a meeting with finance.
The implication for how you use AI day to day is straightforward. Let it do the tasks. Use the time you save to be in more conversations with the business. That’s where your value lives, and it’s the part the robots cannot touch.
AI and Business Partnering is About Priority and Balance
The two questions I get asked most by CFOs are:
Should we be investing in AI or in business partnering capability? And which matters more in the next two years?
The answer to both is the same. It’s not an either-or. It’s a polarity.
A finance team with great AI tools and weak business partnering becomes a faster reporting function. Cheap, efficient, and increasingly invisible to the decisions that matter.
A finance team with great business partnering and no AI becomes a slow, expensive, insightful function that can’t keep up with the pace the business needs.
You need both. Read my post on business partnering vs artificial intelligence for more insights on the AND-not-OR thinking that the smartest finance leaders are now applying.
If your function is investing in AI but not investing in the partnering skills that turn AI’s outputs into actioned decisions, you are building half a function. CFOs who implement strategic AI deployment are predicted to add 10 margin points of growth by 2029. This prediction is based on the whole package. Strategic deployment, not isolated pilots. Governance and skills, not just tools. If you want further coaching and training on this, take a look at The Leaders Lab.
A 30/60/90 Day Starter Plan
If you’ve read this far and want a concrete starting point, here is what I’d suggest:
If you’re an individual finance professional:
In the first 30 days, pick one of the six use cases above and use AI for it daily. Just one. Commentary drafting is the highest-leverage starting point for most business partners. Dataset cleaning is the highest-leverage starting point for most analysts. Build the habit before you broaden.
In the next 30 days, add a second use case and start critiquing the output more rigorously.
Ask yourself:
- Does the commentary line up with the numbers?
- Are the drivers real?
- Would I be comfortable defending this in a meeting if challenged?
This is the audit-mindset habit that separates responsible use from blind trust.
In the final 30 days, share what’s working with the rest of your team. The fastest way to consolidate the skill is to teach it to one other person.
If you’re leading a finance function:
In the first 30 days, audit your master data honestly. Pick the three datasets you would most fear running AI against, and put a small remediation project around them. In parallel, identify three to five volunteers across the team (across mixed grades, with a curious mindset) to be your first cohort.
In the next 30 days, set the ground rules:
- What AI is allowed to do without human review
- What requires sign-off?
- What is off-limits entirely?
Train the cohort on the use cases that matter most to your function.
In the final 30 days, measure. Time saved per person per week. Specific outputs improved. Number of business partnering conversations the team is now having that they weren’t before. That last number is the one that matters most. If AI saves your team time and that time isn’t going into more business conversations, you’ve automated yourselves into a corner.
The Bottom Line on How to Use AI in Finance
AI is now a default tool in the finance function. The teams that win are not the ones with the best tools or the most sophisticated pilots. They are the ones whose people use AI to clear the noise so they can do the work AI cannot do – see context, exercise judgement, and translate numbers into decisions for the business.
And remember the principle behind everything we write here.
Once the robots arrive, the only thing left will be business partnering.
The point of using AI well isn’t to become an AI expert. It’s to give yourself the time, the headspace, and the credibility to be the finance person the business actually wants in the room.