There is a side to accounting that rarely makes it into the recruitment brochure.
Earlier this year, The Wall Street Journal profiled Iowa auditor Matt Gardiner, who regularly climbs 90-foot grain bins to verify millions of bushels of corn. The job can involve fertilizer-soaked boots, dust thick enough to require being hosed down afterwards, and the occasional encounter with pigeons, mice and sleeping bats.
Other auditors interviewed for the piece described counting frozen fish in industrial freezers, rocks in snake-infested quarries, pigs on farms and thousands upon thousands of nuts and bolts.
You’d think this would be precisely the sort of work technology would have eliminated by now. Not quite. Drones have helped with some aerial inventory counts and AI is increasingly being used to analyse inventory data and identify higher-risk areas. But regulation, cost and the awkwardness of the physical world mean human auditors are still required to do plenty of counting themselves. Drones struggle indoors and with obscured objects, while US auditing rules have historically required physical verification of inventory.
It’s a wonderfully messy metaphor for what is happening to accounting more broadly. AI is advancing extraordinarily quickly. But the future of accounting isn’t arriving evenly, and understanding what technology can’t do may become just as important as understanding what it can.
AI has crossed an important threshold
For years, accountants could reasonably look at generative AI and think: impressive, but not quite ready for us. Large language models were terrific with words but much shakier with numbers. They could draft an email or summarise a report beautifully, then make the sort of numerical mistake that would make an accountant break into a cold sweat.
That gap is closing. A 2025 study by researchers from New York University Stern and GoodFin evaluated 23 large language models on the CFA Level III exam, including its notoriously difficult constructed-response questions.
And the progress is increasingly showing up in practice. Late last year, PwC Chief AI Officer Scott McLiver described a simple test he’d been running since the GPT-3.5 era. He would give each new model a 50-line trial balance and ask it to produce basic financial statements. Every previous model had failed his test, often stumbling over seemingly straightforward issues such as accumulated depreciation. GPT-5.2, he reported, got it right.
That’s one practitioner’s informal experiment, not a controlled study. But it captures something the broader evidence is increasingly confirming.
A systematic review published in late 2025 examined 575 academic papers on AI in accounting and auditing between 2021 and 2025. It found AI already being applied to everything from fraud and anomaly detection to financial forecasting, audit accuracy, reconciliations and report generation.
Research interest itself has exploded. The review identified roughly 63 relevant publications in 2021 and nearly 247 in 2025.
But perhaps the more telling figure comes from the profession itself. Thomson Reuters’ 2026 research found that 81% of tax and audit firm professionals are now regularly using AI in their day-to-day workflows. Another study spanning more than 1,500 professional-services workers found organisational use of generative AI had almost doubled in a year, from 22% to 40%.
AI in accounting is no longer an interesting experiment. It’s becoming part of the job.
The most important question isn’t which jobs disappear
Naturally, this has revived predictions about accountants becoming obsolete. I’m not convinced that’s the most useful way to look at it.
Technology rarely eliminates an entire profession in one clean sweep. It tends to disassemble jobs into their constituent tasks, automating some, augmenting others and increasing the value of capabilities that suddenly become relatively scarce. That’s exactly what appears to be happening here.
Accounting has always contained an enormous amount of structured, repeatable work. Data entry. Reconciliation. Classification. Document review. Standardised checks. First-draft reporting. Basic research. These are precisely the environments in which AI and automation thrive.
The systematic review found that RPA and AI are already reducing repetitive workloads, while more sophisticated systems are improving anomaly detection and analysis. It also points towards a profession moving from retrospective data processing towards more predictive and proactive work.
That doesn’t make accountants less valuable. But it does change where their value comes from. And I suspect this is the shift that matters most.
For much of the profession’s history, knowing how to produce the numbers was valuable partly because producing them required expertise, time and labour. As more of that process becomes automated, producing the information becomes less differentiating.
Understanding what the information means becomes more important. So does knowing whether it can be trusted, what questions to ask next and what someone should actually do about it.
The junior accountant problem
There is, however, an uncomfortable consequence buried in all this. How do you develop senior judgement if AI starts doing the junior work?
For generations, professional expertise has been built through repetition. Young accountants reconciled accounts, reviewed documents, prepared first drafts and performed relatively mundane audit work. It wasn’t glamorous, but in the process they developed pattern recognition, technical fluency and an instinct for when something didn’t look right.
Some of that work is precisely what AI is now absorbing. Thomson Reuters found that 49% of professionals working in tax and audit firms expect entry-level roles to decrease over the next two to three years.
AICPA has become concerned enough about this issue to launch its Profession Ready Initiative. Its 2026 research argues that automation, AI and changing business models are reshaping many of the entry-level tasks through which new accountants traditionally developed experience and judgement. Technical accounting knowledge remains essential, but early-career professionals increasingly need to connect that knowledge to business context, communicate clearly, exercise judgement with technology and adapt as expectations change.
This creates something of an apprenticeship paradox. We want AI to take away the grunt work. But grunt work was sometimes where expertise was born.
The answer surely isn’t to preserve inefficient work simply because that’s how previous generations learned. But firms will need to become far more intentional about creating the experiences through which judgement develops.
You can’t automate the bottom rung of a career ladder without thinking carefully about how people reach the second rung.
Clients aren’t asking for less. They’re asking for more.
There is another reason I don’t buy the simplistic ‘AI will kill accounting’ narrative. As technology improves, client expectations are rising with it.
Thomson Reuters found that 89% of corporate tax clients consider AI-enabled quality improvements from their professional firms very important or essential. At the same time, its research shows significant pressure on firms from clients expecting them to move faster.
But clients aren’t simply asking their accountants to process the same work more efficiently. They increasingly want advice. The 2026 State of Tax Professionals Report found 74% of respondents said their clients strongly wanted additional financial advice. Firms are consequently expanding into areas such as tax strategy and business consulting, both as sources of growth and as ways of differentiating themselves.
This makes sense. A business owner rarely lies awake at 2am wishing their accountant would reconcile a ledger more efficiently. They’re worrying about whether they can afford to hire another person. Whether cash flow will hold up. Whether an investment makes sense. What a regulatory change means. Whether they’re missing something. What they should do next.
The accounts are important because they provide evidence. The value comes from helping someone make a better decision.
Perhaps the future accountant looks less like a calculator
There is an interesting clue to this future in a trend that, at first glance, has almost nothing to do with AI. Gen Z has started throwing tax parties.
In the US, informal ‘admin nights’ began appearing on TikTok as young people gathered with friends to tackle unpleasant life administration together. Tax software provider TaxAct subsequently formalised the idea as National Admin Night.
In a TaxAct-commissioned survey of 2,000 US adults, half said they’d file their taxes earlier if somebody else were doing it at the same time. Among Gen Z, 41% said tackling an administrative task alongside someone else would help them follow through, compared with just 14% of Boomers.
There are obvious limits to what we should infer from a commissioned US survey about a promotional campaign. But I think the behaviour is interesting.
For all our sophisticated technology, people still want someone beside them when something feels complicated, uncertain or consequential. That’s worth remembering as we imagine the accountant of the future.
Karbon captures the idea neatly in its framing of AI and accounting service: AI provides speed, while people provide empathy and personal connection.
I wouldn’t draw the line quite that neatly because AI will undoubtedly become better at interactions that feel empathetic too. But trust, accountability and human judgement matter enormously when the stakes are high.
The academic evidence reinforces this. Researchers have identified objectivity, confidentiality, transparency, accountability and reliability as critical ethical dimensions in AI-enabled accounting. The conclusion isn’t that AI has no role in judgement, but that effective systems require human oversight and clear responsibility for consequential decisions.
The accountant of the future may therefore spend less time acting as a human calculator and more time as an interpreter, adviser, questioner and trusted second set of eyes.
AI may also break the business model
There is one more uncomfortable implication. If AI makes accounting work dramatically faster, what happens to a profession that has traditionally charged for time?
Imagine a task that once took five hours can now be completed, reviewed and delivered in one. That’s terrific for productivity. It’s rather less terrific if your revenue model depends on selling five hours.
Hourly billing is still dominant. Thomson Reuters’ 2026 research found 78% of surveyed firms primarily use it. Yet the same research points to growing interest in value-based pricing as automation reduces the relationship between hours worked and value delivered.
This could turn out to be one of AI’s more significant impacts on professional services. The technology won’t merely change how accountants work. It could change what firms sell.
If clients can increasingly access processing, compliance and analysis cheaply and instantly, charging for the mechanics becomes harder to defend. Firms may need to monetise something different: interpretation, judgement, reassurance, strategic insight and outcomes. After all, AI technology can tell a client what they COULD do, but it takes uniquely human skills and judgement to be able to confidently advice what a client SHOULD do.
In other words, the unit of value may gradually shift from work performed to confidence created.
The winners won’t necessarily be the firms with the most AI
There’s a temptation whenever a technology moves this quickly to assume the winning strategy is simply to adopt more of it. The evidence suggests otherwise.
AICPA and CIMA surveyed more than 1,400 finance leaders and found 88% believed AI would transform the profession within two years. Yet just 8% felt very well prepared to adopt it, while half identified a shortage of skilled talent as their biggest obstacle to adopting new technologies.
Even more strikingly, Thomson Reuters found 35% of professionals in tax and audit firms were using AI their organisation hadn’t approved. That’s not really a technology problem. It’s an organisational one.
The firms that adapt well will need to rethink workflows, training, governance, pricing and career development together. They’ll need to decide which work machines should do, which work people should own and where the combination produces something neither could deliver as effectively alone.
They’ll also need to resist one of the easiest mistakes to make with AI: using a genuinely new technology to make an old operating model slightly faster.
When the numbers get easier
Back in Iowa, Matt Gardiner is still climbing grain bins. And somewhere else an auditor is probably standing in a freezer counting cartons, wondering why artificial intelligence can pass sophisticated professional exams but hasn’t yet found a way to save them from frozen fish.
That’s the strange thing about technological revolutions. The future rarely arrives in the order we expect. Some apparently sophisticated tasks disappear surprisingly quickly. Some supposedly simple ones stubbornly remain. Accounting will be no different.
AI will almost certainly become better at processing financial information, identifying anomalies, preparing reports, conducting research and performing work that once required significant amounts of junior professional time. Some roles will shrink. New ones will emerge. Career paths and firm economics will change.
But that doesn’t necessarily diminish the accountant. It changes the basis on which accountants earn their value.
When producing the numbers becomes easier, knowing what they mean matters more. When information becomes abundant, judgement matters more. And when clients can get answers almost instantly, having someone they trust to help them decide what to do with those answers may become more valuable still.
Perhaps that’s the more useful way to think about the future of accounting. Not what happens when AI learns to do the accountant’s work, but what becomes possible when accountants no longer have to spend so much of their time doing work a machine can handle.
Michael McQueen is a globally recognised trend forecaster, change strategist and keynote speaker.
A bestselling author of 10 books, Michael’s latest release was named by Malcolm Gladwell and Adam Grant as one of the top five must-read new leadership books. He is a sought-after media commentator, with his insights regularly featured in Forbes, The Guardian, and CNN.
To find out more about Michael and his work, click here.
