The work AI does best is the work nobody wants to do. In finance, that is precisely the problem.

For decades, junior accountants learned their trade by reconciling accounts, manually entering data, and grinding through the month-end close. The tasks were tedious, but they built an understanding of the mechanics of finance. It’s now possible to automate much of that work, but doing so removes a crucial stage of training for future finance chiefs.

“I really do worry how the next generation of controllers, finance directors and, eventually, CFOs will learn when those entry-level jobs are not the same as what they used to be,” says Catherine Birkett, CFO of London-based payments company GoCardless. An understanding of the data can help finance professionals get more out of AI tools, she argues, and this is often developed in junior finance roles.

“A year ago finance chiefs were unsure what difference AI would make. Now every single one of them is using it,” says Mark Freebairn, head of the board and CFO practice at executive search firm Odgers. He predicts that AI could eventually automate much of the accounting function, adding: “If it’s not replacing all of it yet, it will do.”

A diamond-shaped function

As the use of automation increases, the number of entry-level finance roles is shrinking. PwC took on 200 fewer entry-level staff in the U.K. last year, and nearly two-thirds of CFOs surveyed by the Oliver Wyman Forum and the New York Stock Exchange expect junior roles to make up a smaller share of their finance teams over the next three years.

“There are absolutely going to be fewer entry-level finance jobs,” Birkett says. “Those that remain will go to capable candidates who look little like the accountant of 30 years ago,” she adds.

With companies seeking to cut the number of junior hires in the finance function, the traditional pyramid-shape of the function is changing. Many companies are moving to a “diamond-shaped” structure, with more mid-level finance professionals and fewer trainees, according to Freebairn. But this presents new problems for the profession.

“If you want a hundred managers but you only need two trainees, when those managers have retired and you’ve only got two people to replace them with, what are you going to do?” he asks. The usual answer is to poach from rivals. “If you’re all creating diamond management structures,” Freebairn adds, “there won’t be anyone to steal.”

“Nobody’s worked it out yet,” Birkett says, adding that the Big Four audit firms, which hire graduates by the thousand, must be wrestling with the same dilemma.

With fewer trainees at the start of the talent pipeline, organizations are likely to face a thinner shortlist for the CFO seat in the future. This change comes at a time when boards are already facing challenges filling the role. CFO turnover is at a record high and many companies are unprepared for the level of churn—only 16% of finance chiefs say their organization has a proactive succession plan, according to Russell Reynolds’ Global CFO Turnover Index 2025 Report.

Christophe Catoir, president of the recruitment group Adecco, warns against experienced professionals “pulling the drawbridge up after them.” Those hit hardest, he says, will be young people without professional networks or the means to take unpaid work. This may risk undoing progress on diversity in finance. “Technological progress should be used to help the workforce, not become another barrier to social mobility.”

Not all finance firms are cutting junior headcount. Brian Spaans, CFO of Bunq, the Amsterdam-based digital bank, is still hiring graduates and sees AI as “an opportunity to develop better finance professionals.” With machines handling the repetitive work, graduates can start analysis far earlier, he says.

Rewriting the interview

The increased use of AI in the finance function is also impacting the skills that companies assess for when hiring for entry-level talent.

Junior finance candidates at GoCardless used to be given a data set and a copy of Excel as part of the final round of testing. Now Birkett gives candidates an AI model and asks them to analyze the data and build a presentation. The point is to see whether they accept the AI’s first answer or push it to dig deeper. “The absolute worst thing that could happen in finance is that people just accept what they’ve been told,” she says.

Catoir warns against assuming that digital natives will naturally be equipped to work with AI. “Being comfortable with technology does not necessarily mean you understand how to use it effectively,” he says.

For Freebairn, the increased use of AI makes technical expertise even more important. “AI is going to require more technical competence than ever,” he says. “Someone has to know when the machine is wrong.”

GoCardless hires graduates from any discipline but still encourages them to qualify as accountants and pays for the exams.

Bunq has gone further, abandoning CVs altogether. Instead, applicants must complete a seven-step process that includes personality and abstract reasoning tests. Successful candidates will then be invited to spend a day at the bank’s offices where they are tested on banking regulation, their use of AI, and their ability to build calculations.

Spaans says the company is looking to assess whether candidates can find the information they need to complete the tasks and apply it correctly. One of his junior regulatory specialists came from data analysis in sales. “If we had been screening CVs in the traditional way, we probably would not have looked twice,” he says. “We are not hiring for where someone has been. We are hiring for how they think.”

Learning the trade

Hiring is only half the problem. Technical knowledge can be taught and examined but the skills Freebairn values most, such as influencing, persuading, and negotiating, cannot. “How do you create that if you don’t have development jobs? Some lessons only come from being in the room, haggling with a buying director over which T-shirts to stock, getting it wrong and doing better next time.”

With AI taking away some of the opportunities for junior accountants to hone these skills, Birkett says: “We’re going to have to accept that juniors will take longer to earn their keep.” In practice, that means juniors sitting in on meetings to watch colleagues interrogate the numbers, being present when real decisions get made, and gaining experience in different departments. It may also mean continuing to do some reconciliations by hand.

At Bunq, junior staff are still expected to do reconciliations to check the data against the bank’s records and to document the process for auditors and regulators.

“You cannot automate a process you do not understand,” Spaans says. “AI speeds up the work. It does not remove the need to understand it. If anything it raises the bar, because when output arrives faster, ownership of whether it is right lands more squarely on the person who built it.”

Spaans argues the approach is working. Staff in Bunq’s finance and treasury teams stay three to four years on average. “In a field where junior churn is a known problem, that suggests the work feels worth staying for,” he adds.

Bunq’s approach to hiring and training is helping it build a strong pipeline of finance talent. Companies that have yet to find an answer, or are simply cutting junior roles, could face a far harder problem when it comes time to find their next CFO.