FP&A and performance management
AI in finance: from first steps to agents
Finance can win back a lot of time for analysis. For that, the data, the boundaries and the controls need to be in order.
Few functions have as much to gain from AI as finance. The work revolves around numbers, comes back every month and involves a lot of documents. At the same time, finance needs exact figures that can be checked, while a language model gives the most likely answer. You have to take that tension into account when you start.
Start with the work around the numbers
The first gains usually come from the work around the calculation: a first draft of a commentary, a summary of a long document or the preparation of a review. You see the result straight away and the professional stays in control.
Decisions, journal entries and anything that goes to the board or the auditor remain human work. AI drafts, the professional reviews and signs off.
Treat AI output as a draft
Language models write convincingly, even when a number is wrong. So treat the output as the work of a new colleague: often useful, but only final once someone who knows the numbers and is accountable for them has looked at it. That only works with clean, well-governed data. On a weak foundation, AI makes errors bigger.
The next step: agents
Salvatore Cantale of IMD describes agentic AI as software that pursues goals on its own: planning, reasoning, using tools and adjusting course. He compares it to a digital junior controller and cites Gartner, which expects around 15% of routine work decisions to be made autonomously by 2028.
For finance, Cantale sees three levels. Task agents carry out work, such as processing invoice data or reconciling transactions. Workflow agents coordinate, for example the information flows in a budget cycle. Strategic agents act as a sounding board, with what-if analyses on currencies and interest rates.
His advice to CFOs: bring the data together first, set out how far an agent's autonomy goes and when it must escalate, and require agents to show their reasoning, especially in reporting and tax.
Governance and people
Arturo Bris advises CFOs not to wait for regulation. Build a central overview of every AI tool in use, assign owners and train staff in responsible use. Focus reskilling where AI adds the most value. And pay attention to the people: many finance professionals are wondering what AI means for their own role.
Measure it
Agree up front how you will measure success, for example in lead time, error rate or time freed up for analysis. Have the business owner confirm the result. Only then can an experiment become a business case.
Sources
- Bris, A. (2025). AI and the future of finance: How CFOs can navigate diverging regulations and workforce transformation. I by IMD, 17 March 2025.
- Cantale, S. (2025). Making agentic AI work: What CFOs need to know. I by IMD, 29 October 2025.
Read next
Let's talk
Would you like to exchange ideas on data, analytics, AI or FP&A? I would be glad to connect.