When most industries go looking for AI, they have a demand problem: they want more customers, or cheaper ones. Accounting has the reverse. The customers are already there. The people to serve them are not.
In a global survey of about 500 accountancy leaders, 73% said they were turning away prospective clients because they did not have the staff to take them on. That is an unusual thing to read about any business. It means the ceiling on growth for a large share of firms is not marketing, pricing or reputation. It is hours.
Why hiring does not close the gap
The obvious answer is more accountants, and the pipeline is finally moving the right way. Accounting enrollment at US four-year undergraduate programs rose 8.9% in spring 2026, the third annual increase in a row. The Bureau of Labor Statistics projects roughly 115,300 openings for accountants and auditors every year through 2035, many of them created by people retiring or leaving the field.
Put those two figures side by side and the timing problem is obvious. A student who enrolled this spring is years away from being useful to a four-person practice, and the openings keep arriving in the meantime. In a small firm the shortfall looks familiar: the owner doing staff work at night, the senior who has not had a real vacation since the last busy season, the prospective client told to try again next year.
The work is the right shape for AI
What turns this from a staffing story into a technology story is what those missing hours are actually spent on. Bank feeds waiting to be coded. Statements that have to be tied out. The fourth email asking a client for the same missing 1099. Data rekeyed from one system into the next because the two never learned to talk. None of it is trivial, and all of it has to be right, but very little of it needs an accountant’s judgment on every line. Even the BLS outlook for the occupation anticipates automation absorbing that routine layer while the role tilts toward analysis and advisory work.
On paper, then, this is one of the easiest markets AI could ask for: a profession short of capacity, with a large share of its time spent on repeatable work. Give existing teams back even part of those hours and firms can serve the clients they are currently sending away — a conversation about growth rather than about cutting jobs.
Where the software playbook stalls
The familiar way to deliver that is the one the last twenty years taught everyone: build a better product, sell it by subscription, and wait for firms to adopt it. Two things get in the way.
Building is no longer the hard part. AI makes software cheaper and faster to write for everyone, competitors included. A vendor that ships a genuinely better intake or reconciliation tool this quarter should expect a credible imitation not long after. When code stops being scarce, the product stops being the moat. What stays scarce is getting technology into daily use inside real firms.
Firms rarely replace their core systems. A practice-management platform is not one tool. It carries a decade of client files, the billing setup, who can see what, every integration the firm depends on, and the habits the whole staff has built around it. Swapping it out means migration risk, retraining and weeks of distraction, all scheduled around deadlines that do not move. A vendor’s roadmap moves in weeks. A firm’s decision to change the system everything runs on comes along a few times in a career. And the firms that would gain the most — the understaffed ones — are exactly the ones with no spare hours for product demos in February.
The result is a widening gap between how fast the technology improves and how fast the profession can absorb it. Better software, on its own, does not close that gap.
Ownership as the delivery channel
The route I think gets underrated starts from a different moment in a firm’s life. Tens of thousands of owners are going to step back over the next decade whatever happens with AI. Every one of those transitions is already a reset: a new person signs the engagement letters, looks at how the work gets done, and decides what to keep. If that person arrives with the technology already built and already running in other practices, the adoption problem mostly disappears. There is no pitch to sit through, no migration to fit around filing season, no owner being asked to bet the firm on a vendor’s promise. The question stops being “will this firm buy the tool?” and becomes “how carefully will the new operator bring it in?”
That second question matters more than it sounds. A tool bolted onto a patchwork process usually adds a step. When the people who own the process can rebuild it from the start, the tool can take several steps out instead. That is the difference between staff learning another login and staff getting back the afternoon they used to spend on reconciliations.
For clients, the right amount of visible change in the first months is close to none. At Numica we keep the same people on the relationship and the same rhythm of communication, and change the machinery gradually behind them: documents that arrive without a chain of reminders, a close that happens days sooner, and a March that stops being the month everyone dreads. It is also why I have argued that what AI actually changes when you sell a practice has much less to do with headcount than most sellers fear.
What this does not prove
None of this makes an acquirer a technology company. Buying firms is a financial skill; running software well inside them is a separate one, and plenty of buyers have the first without the second. That is why it is worth making any buyer show you what is actually different at the desk before you believe the word “AI” in their pitch. And the vendors are not going away: firms that stay independent will keep buying good tools, and should.
The claim is narrower than “acquisition beats software.” It is that the slowest part of bringing AI to accounting is not writing the code but getting it adopted, one firm and one owner at a time, and that the profession already has a mechanism that resets thousands of firms at once. Firms would be changing hands with or without AI; the succession math made that inevitable years ago. What AI adds is a reason for those transitions to mean more than a new name on the door — and, for the owner deciding who takes over, one more thing worth asking every buyer about.