Where AI is useful in a finance function today
The useful applications are less glamorous than the demos. They are the hours of reading, sorting, and first drafts that fill a finance team's week.
Most of what a finance team does in a week is not calculation. It is reading contracts, sorting transactions, chasing explanations, and writing the first draft of something. That is where current AI tools help. It is a narrower claim than the one in most vendor presentations, and I think it is the more useful one. I have worked with large language models since 2022, when I led the investment in a product built on them, and I now put these tools to work across a portfolio of companies, so I have seen what they do well and what they cost.
Reading at volume
Every company has more contracts than anyone has read recently. Renewal dates, payment terms, price escalators, and termination rights are buried in them. A tool can pull those terms from several hundred agreements in an afternoon, and a person can then read the twenty that matter.
Vendor consolidation, in my experience, starts here. You cannot renegotiate what you have not found.
Sorting spend
Classifying vendor spend by category is tedious and rarely finished. Done well, it shows the duplicate subscriptions and overlapping tools that accumulate across teams and, in a portfolio, across companies. This is pattern-matching work, and software is good at it.
First drafts
The monthly explanation of why a cost line moved, the memo summarizing a contract, the answer to a question a business leader asks every quarter: an analyst who edits a draft finishes sooner than one who starts from a blank page. The analyst still decides what the draft says.
Where it does not help
Judgment. A tool can tell you that a contract renews in March. It cannot tell you whether to renew it, what to ask for, or how much the relationship is worth. Negotiation, prioritization, and deciding which assumption in a plan to believe remain human work.
How to start
Pick one recurring task that takes at least a day a month. Measure the time before and after. Keep the person who did the task in charge of it. A handful of small, measured wins does more for a finance team than a transformation program with a name.
The longer question, what a finance team is for once this work costs almost nothing, is the subject of When the numbers get cheap.