Breaking the Isolation of Expertise: Why Finance Needs People Who Speak Both Accounting and AI
Updated: Oct 12
For most of my career, expertise
has been organised in separate rooms. Finance sits in one, data and technology in another, and the two meet when a project forces them to.
That arrangement worked while technology changed slowly. It does not work now. AI can read invoices, match transactions and draft commentary, so the hard part of automating finance is no longer the tool. The hard part is finding people who understand both the accounting and the technology well enough to trust the result.
Why the crossover matters more than ever
Gartner's 2025 survey of 183 CFOs and senior finance leaders found that 59% of finance functions were using AI. That was almost unchanged from 58% a year earlier, after a jump from 37% in 2023. The same survey named data literacy and technical skills, alongside data quality, as the largest obstacles.
The technology kept improving over that year, so the brake is somewhere else: in skills and in data. The tools are ready. The combination of knowledge needed to use them safely is scarce.
What each discipline misses on its own
A data scientist without finance. Will build a matching model that is right, say, 97% of the time and call it a success. In finance, the other 3% is where the audit finding sits. Without knowing what materiality, cut-off or segregation of duties mean, it is hard to tell which errors matter.
An accountant without data skills. Either avoids the tools or buys them without being able to question the output. Both are risky. You cannot review what you do not understand, and review is the control.
Both, working separately. Produce a long handover document and a system that nobody fully owns.
What the crossover looks like in practice
The people who make finance automation work are translators. They can read a trial balance and a data pipeline. In practice, they:
design the control into the automation from the start, with an input, a rule, an output and a reviewer
know which process to automate first, because they have done it by hand
ask whether a result can be evidenced before asking whether it is clever
explain the outcome to an auditor and to an engineer in terms each will accept
I came to this from both sides, as a finance director who went on to study AI after several digital transformations. Neither discipline alone would have been enough for the work I do now.
Building it in your team
You do not need to hire a new kind of person. You can grow the crossover in the team you have.
Pair people on one real process. Put an accountant and an analyst on a single reconciliation for a month. Each learns more from the other's questions than from a course.
Take the data team through a close. Sitting through a month-end shows them where the pressure and the risk are.
Fund the learning. A finance analyst who learns SQL, basic Python or how to write and test a prompt becomes more valuable quickly. Give them time in working hours to do it.
Reward the crossover. Recognise the people who work across the boundary, in reviews and in promotion decisions, so that it is seen as a route up.
Run short knowledge sessions. A regular lunch-and-learn, where finance explains an accounting concept and data explains a technique, costs almost nothing.
A culture that allows the unknown
None of this works where people are afraid of looking ignorant. An accountant has to be able to ask what a model is doing, and an engineer has to be able to ask what an accrual is.
Leaders set that tone by asking basic questions themselves, and by accepting that some experiments will not pay off.
Measuring whether it is working
Measure the crossover by what it delivers:
hours taken out of the close, measured before and after
the share of automated output that needs manual correction
the number of processes with a documented, auditable automated control
how many people in finance can work with data tools, and how many in data understand the close
Final thoughts
Specialist depth still matters. What has changed is that depth in one field is no longer enough to capture what AI offers finance.
The organisations that benefit will be those that stop treating accounting and data science as separate rooms. The people who benefit most will be those who are comfortable in both.
Comments