First in a series on automating an accounting firm. The series will take a firm apart process by process and rebuild the pieces; this first post is about what that does to the firm itself — its prices, its size, and the people inside it.
On the morning of 1 May 1975, a New York stockbroker went to work for a business that no longer existed in the form he had joined. The Securities and Exchange Commission had ended the fixed commission — the rate every member of the New York Stock Exchange was obliged to charge and forbidden to undercut, a floor under prices that had held in one form or another since traders signed the Buttonwood Agreement under a tree in 1792. The industry had fought it to the last, lobbying, threatening to sue, and warning that competition would be its ruin; the very nickname, May Day, borrowed the distress call of a sinking ship. Overnight, commissions on large institutional orders fell by around half, and kept sliding in the years that followed. By the end of the year thirty-five brokerage firms had gone. There were stories of men who had earned six figures in the spring and were driving cabs by the winter.
Broking itself carried on. The casualty was narrower: the broker who only took the order. Executing a trade was the commodity — structured, repetitive, identical at every firm — and once the protected price came off it, doing it well distinguished nobody and paid almost nothing.
The survivors had something the wire could not carry: advice, judgement, a relationship clients paid for because they wanted it. The discount brokers took the commodity and the volume; everyone else moved up into wealth management or disappeared. The line between the two ran exactly where the machine’s reach ended.
The fixed price nobody calls one
An accounting firm has a fixed price of its own, though nobody files it with a regulator. Watch where the hours actually go and the accounting is the small part. The ledger, the tax computations, the statutory filings are well served already, by software every firm runs and a crowded market keen to sell more of; the clock fills with everything around them — capturing a document, chasing the client who has not sent it, filing it where it can be found again, watching a deadline, reconciling a statement that nearly matches, drafting a report someone will read for ninety seconds.
That operational layer is the firm’s real cost, and it is exactly the work this generation of AI handles well: structured, repetitive, high in volume, bound by rules. None of it is hypothetical. Debitus, a Portuguese firm, already has the first piece running, every incoming document captured, classified and indexed before a person touches it.
The cost of that layer has quietly held the floor under the fee. A firm charges what it charges because serving a client takes a known quantity of human time, and below that the price cannot go. Automate the operations, as this series sets out to do, and the floor falls out.
The tempting assumption is that the firm pockets the difference. It does not, and the reason is structural. Accounting is compulsory: a company must file, must close, must comply, whether it wants to or not.
Demand that cannot escape gets bought the way people buy anything they are forced to buy — on price.
The moment one firm can serve a client for half the cost, the half-price offer appears, and the client, who never enjoyed the bill, takes it. The saving does not settle in the firm’s margin; it is competed down the chain until it reaches the client as a lower fee. I wrote in an earlier post about the value you can’t keep — value a technology creates that nobody deploying it manages to hold. Compliance, once it is automated everywhere, is the cleanest example I know.
Two ways to spend a saving
This puts a real decision in front of the firm that automates first. A firm that has halved its cost to serve can do one of two things with the saving:
- It can shrink to fit the new price — fewer people, the same clients, a smaller version of itself.
- Or it can hold its structure and spend the freed capacity taking its rivals’ clients at a price they cannot match.
The arithmetic of the second is hard to ignore: a firm built to carry a hundred clients can, with the operations automated, carry five hundred or a thousand on much the same staff. Volume climbs, margin per client falls, and the firms that did not automate watch their book walk out of the door.
None of this is a forecast. It is the present, and it has a balance sheet behind it. Private equity worked out before most firms did that a fragmented, cash-generating, recurring-revenue profession is a consolidation waiting to be funded. The pattern is a roll-up: buy one firm as a platform, then fold dozens of smaller ones into it. The International Federation of Accountants counts fewer than two hundred direct private-equity investments that have drawn in nearly nine hundred further firms, and reckons each direct deal now triggers about seven and a half others — roughly four times the rate of 2021. The annual deal count tells the same story in plainer terms.
The retainers are going the way of the commissions. Compliance pricing is sliding towards commodity, and the firms competing on it alone are already watching their margins thin — a point the software vendors selling them the automation are happy to make, on the way to selling them the next module.
The people, and why 1975 is unkind here
Accounting has run this triage before, in miniature. The comptometer operator — a real, skilled and respectable job of mechanical calculation — did not survive the electronic calculator. The machine has thinned these benches once already.
For most of a firm’s staff the operations were not a part of the job; they were the job. Capturing, filing, chasing, reconciling — that was the day. When the day’s work is automated, the role empties, and for most of the people in it there is nothing more interesting waiting underneath to refill it. A firm can move some of them up into the work that is left — the exceptions, the advice, the conversations a client wants a person for — but fewer will make that crossing than the retraining slides assume, and it is worth being straight about why.
Advisory work rewards a tolerance for ambiguity, a willingness to be wrong in front of a client, and a pull towards the client’s problem and away from the form — close to the reverse of what three decades of compliance work selected and rewarded. Some of the best operations people chose that work precisely because it was bounded, ruled and free of surprises; the firm sold them that bargain, and automation breaks the firm’s side of it. A practice that budgets to turn its whole staff into advisers is budgeting against the base rates. The honest plan moves a few, and is candid with the rest. It meets a profession already thinning at the entrance — fewer people sitting the exams, a retirement wave pulling out the senior ranks, a trend the AICPA’s pipeline data has tracked for most of a decade — and thins it again in the middle.
What the firm actually keeps
So where does anything stay? In the part that was never operational. When every firm can capture a document and hit a deadline, those things no longer command a premium, and the premium moves to what cannot be lifted off a shelf: the firm’s own governed data, built over years and trustworthy enough to act on; the processes it has shaped around its own clients, not taken off a vendor’s shelf; the particular choices it makes about where to point AI and where to keep a human; and the judgement sitting on top of all of it. That is the margin the automating firm keeps. The cost saving leaks; this does not. It is also, not by accident, the only ground on which a person in the firm is worth more than the model.
Two questions sit under everything that follows in this series, and both are managerial long before they are technical. Where do the people go, once the work that filled their day is done by something else? And how does a firm still stand apart when the basics it used to bill for are free to everyone? Over the next posts I will take a firm apart a process at a time — documents, deadlines, the client relationship, the close and reconciliations, reporting, the back office, hiring and training — and, where it earns the space, build a working piece of each, so you can watch the thing run instead of taking my word for it. The destination is set from here: a firm that does less accounting and more of the work around it, charging less, carrying more, and keeping only the part a machine cannot copy. The managing partner’s real work for the rest of the decade is deciding, early, what that part is — before a competitor at half the price decides it for him.
Most businesses are already facing this same question, just with a different machine. If you’re working out where your own AI agents belong on that grid — and where a human still has to sign — talk to us.
