At a glance
- AI may reduce jobs for young workers but increase them for experienced ones.
- This is because experience complements AI, which can replace knowledge learned from books.
- With AI doing routine tasks, juniors may struggle to gain necessary experience.
- Accountants should focus on selling judgment and teaching it to their juniors quickly.
Readers of Financial Accountant know that artificial intelligence is reshaping accounting practice. The next question is: what will that do to the lives of today’s accounting professionals?
An illuminating answer has now come from Stanford’s Erik Brynjolfsson, one of today’s leading thinkers about the ways that information technology changes our economy.
Brynjolfsson and his colleagues argue today’s experienced accountants may be particularly well placed to profit from AI. But that, they suggest, may be bad news for accountants just starting out.
How AI affects workers by age
Brynjolfsson, along with colleagues Bharat Chandar and Ruyu Chen, recently published a paper called “Employment Effects of Artificial Intelligence” in the prestigious Quarterly Journal of Economics. The three economists made use of some remarkably up-to-date monthly figures on more than three million US workers. (The figures came from the US’s largest payroll-processing company, ADP.)
They used that huge dataset to lay out several tentative findings about AI:
- Early-career employment declined in AI-exposed occupations. Workers aged 22–25 in fields like software development and customer service saw sharp drops in employment after late 2022. Meanwhile, older workers and less AI-exposed jobs remained stable or even saw employment grow.
- Overall employment grew, but not for young workers. While US employment for older workers went up by 6 to 9% between late 2022 and July 2025, 22–25 year-old workers saw jobs drop 6%.
- Automation-linked AI drove job falls. Where AI is mostly used to substitute for people, jobs went down. In jobs where AI complemented human work, it did not; in some such areas, the number of entry-level jobs even rose.
- The big effects happen in job levels, not pay levels. One of the most durable findings of economics is that the price of labour usually drops very slowly. So it is here: according to Brynjolfsson’s team, employers are mostly cutting jobs rather than pay.
The surprise finding on older workers
The most obvious explanation for these results is that the disappearance of some routine work is pushing firm decision-makers in many parts of the economy to hire fewer junior workers. In accountancy, this could mean that firms are finding they can produce the same set of accounts or the same tax return with less work – and so they are deciding to lower their fees, setting new and lower price expectations for clients.
One surprise from this research is that the ADP data’s general finding of stagnant job growth for younger workers is driven by a fall in jobs in AI-exposed industries, like software and customer service. In less AI-exposed industries, such as health care, younger workers are still being hired.
But perhaps the biggest surprise in the research comes in the results for older workers in AI-exposed industries. Their employment has been growing.
Few people had predicted that the data would show AI prompting employers to rush out and hire a bunch of 40-year olds. (Indeed, some early research suggested AI could most heavily affect higher-income workers, who tend to have more experience.) Yet that seems to be what the figures are suggesting.
It may of course be that this result is not connected to AI exposure in the way Brynjolfsson, Chandar and Chen suggest it might be. Indeed, the three researchers themselves caution that they can’t be certain AI explains all the data. We’ll need to wait for further research before we can have any certainty. But already a study using data from LinkedIn and Revelio Labs on 62 million workers has come to similar conclusions.
“To be productive with AI, they need to obtain skills and experience … except that they may not have the opportunities to ‘do.’”
Joshua Gans, economist, University of Toronto
A possible explanation: experience matters for AI
In their paper, Brynjolfsson and his colleagues suggest one good reason why companies might be hiring older workers in AI-exposed fields. This is the idea that AI mostly replaces “codified knowledge”, the type that is written down in books and learnt in university courses. At the same time, it complements “tacit knowledge”, the type you accumulate with years of experience. This uncodified experience of older workers complements the speed and endless energy of AI engines.
In short, AI is working like a team of hyper-enthusiastic interns – cheap, tireless and never late to work. It is easy to see how such a team might benefit from an experienced chief who can give them their tasks.
Brynjolfsson in particular has long argued that technologies like AI gain power when combined with human smarts and insight. Software can flag inconsistencies; only a human can weigh them in context. Clients still crave reassurance, interpretation and advice.
Caution: these are early findings
As Brynjolfsson and his colleagues point out, these are early and tentative findings. Despite competing claims, we still don’t know for sure just how accounting will evolve in the AI age. Many people worry that an AI job-pocalypse is coming to wreak havoc on the entire economy.
People have thought this in the past too, about many other innovations. They have usually been wrong: by creating new and higher incomes, innovation eventually creates more jobs than it destroys.
Still, most economists usually say that innovation can leave some groups of people worse off. Automated weaving really did make life tougher for hand-weavers, just as the Luddites of 19th-century Yorkshire feared.
A generational challenge
The deeper challenge seems to be generational. If the old ladder of apprenticeship collapses – no more years of reconciliations and data entry – how will new accountants learn their craft? University of Toronto economist Joshua Gans puts the dilemma of young workers this way: “To be productive with AI, they need to obtain skills and experience that appear to come from learning by doing, except that they may not have the opportunities to ‘do.’”
There is some cause for hope: another 2025 Brynjolfsson study with two other colleagues suggests AI helps least experienced workers the most – seemingly the opposite of the findings in this latest study. We’ll have to wait to see which view, if either, is closest to reality.
Until then, Brynjolfsson’s work does offer accountants tentative prescriptions:
- Stop selling time; start selling judgment.
- Invest enough in AI to remain credible, but remember that the human connection is what keeps clients.
- Teach juniors judgment quickly; don’t teach them bureaucracy.
- Guard your niche. Context, local knowledge and trust are the moats machines cannot cross.
AI is not abolishing accountants. But it is challenging the profession to think about the way it works.
Join delegates from across the globe to hear from respected industry specialists at the IFA’s International conference online 2025. You’ll hear the latest developments on topics such as AI and international tax. Register now.
