Who Wins the $650BN Accounting Market?

The $650 billion accounting market is up for grabs... quite literally. KKR and OpenAI are buying their way in, and SoftBank is reportedly writing a cheque to join them. Pay-to-win, and accounting is only the first market on the list... the same land grab is coming for legal, insurance and consulting.

One accountant at a firm in Washington State spent 180 hours on tax preparation last year. This year, the same work took her 15. She spent part of the difference calling every client to walk them through their returns... service that was physically impossible twelve months earlier... and the rest taking on new ones.

Her firm did not buy that software. Her firm was bought... by a roll-up vehicle whose engineers built the software in-house, sitting next to OpenAI researchers who are embedded inside the business, because OpenAI owns a piece of the buyer.

That is a 12x collapse in the labour behind regulated professional work... in one year, in production. This edition is about the machine that did it, who is funding it, and what it means for everyone who sells expertise by the hour.

Last week I mapped legal, and every services market resolves into the same positions: incumbents who build their own AI, natives who enter and do the work itself, tools that must eventually choose between selling software and becoming the firm, and a squeezed middle that becomes acquisition inventory.

Accounting answers a question legal could not. In legal, the middle is inventory with no buyer yet. In accounting, the buyer has arrived... with a chequebook, and with a model lab on its cap table.

The numbers first. The global accounting services market is worth roughly $650 billion. The software sold into it... the ledgers, the ERPs, the tax engines... is worth roughly $25 billion. For every dollar spent on the tools, about twenty-five go to the work itself. Legal ran thirty-to-one. Same lopsided shape.

Three quirks make accounting a different war.

First, the revenue renews by law. Tax filings and audits happen every year because the law requires them. Unlike legal, where each new matter is a fresh purchase decision, most accounting work never goes back out to bid... unless the firm makes an error.

Second, the relationships last decades. Businesses change accountants roughly never. Winning those clients one by one would take ten years. Buying the firm transfers all of them in a single transaction.

Third, and this is the hinge: the data the AI needs lives inside the firm. Not one contract at a time, but the entire financial history of every client... ledgers, bank feeds, prior-year returns, workpapers... sitting in twenty-year-old systems with no APIs. In accounting, acquisition is the integration strategy. You cannot plug into the data. You have to own it.

Now the four armies.

Rillet is accounting's Harvey... an AI-native ERP that has raised over $100 million in under a year, with a $70 million round co-led by a16z and Iconiq. Its customers close their books in days instead of weeks, with finance teams a fraction of the old size. Around it sit Basis, Puzzle and Digits, all arming accountants with agents.

But the tools face something legal's tools never did: entrenched incumbent software. Intuit owns most of the small-business ledger market, and Thomson Reuters is already shipping agentic tax preparation into its customer base. And the deeper trap is the same one I described last week... a tool that every firm can rent raises the floor for every firm at once. It makes nobody different.

The entrants doing the work itself... Pilot running the books as a service for startups, Finally doing SMB finance with AI at the core, a growing wedge of AI-first tax boutiques behind them. Same honest caveat as legal: each is winning a slice, mostly work that was previously done badly or not at all. The wedges are real and compounding. None of them, yet, is the main event here.

At the top, the Big Four... $212 billion in combined revenue... are running their own version of the Kirkland play: pivoting from selling advice to operating clients' functions outright. KPMG is targeting up to a fifth of its US consulting fees from these multi-year managed services contracts. One accounting professor put the reason plainly: AI changed the economics. Running the work used to mean thin margins and armies of people. Agents flipped that.

Below them, the picture flips too. The mid-market is not choosing whether to fight. It is losing the ability to field an army at all: 340,000 accountants gone in five years, CPA exam candidates at a seventeen-year low, three-quarters of practising CPAs at or near retirement, and nine in ten firms saying they struggle to hire. A generation of partners owns firms full of renewing revenue and has nobody to hand them to.

Eager sellers, sticky clients, revenue that renews by law. Read those three again and you can price the market yourself. Enter the army legal does not have.

At the top of the market, private equity is already inside: KKR took a majority stake in Crowe at close to $3 billion. Two weeks later, Reverence bought into Eide Bailly at $1.8 billion. Blackstone owns part of Citrin Cooperman. Half of America's twenty biggest firms now have PE behind them... and the Crowe and Eide Bailly deals both priced at roughly two times revenue. Services multiples, about to be run on software margins. That is the whole trade.

At the bottom, something more interesting than a buyout is happening. Thrive Holdings... the permanent-capital vehicle spun out of Josh Kushner's Thrive Capital... has committed $1 billion to rolling up local CPA firms and is now reportedly raising $2 billion more from SoftBank, Altimeter and D1. OpenAI holds equity in the vehicle. Its head of applied research holds a joint role there. Its engineers work inside the acquired firms, not as a vendor but as part-owner.

Its accounting platform, Current, has bought 48 firms. And this spring, the receipts arrived. The Tax AI built inside Current processed 7,000 returns this season, cut preparation time by roughly a third, drafted returns at up to 97% accuracy... and gave that accountant, at Larson Gross, her 165 hours back. This is not a hypothetical: Forbes reported her story, and OpenAI wrote it up on its own blog.

The honest caveat, because announcements run ahead of operations everywhere in this category: that is one tax season, on the simpler end of the return spectrum, with humans reviewing everything. A botched audit is not a UX bug, and nobody has published the margin numbers yet. But it is the first hard evidence anywhere in professional services that buy-the-firm-and-rebuild-it-around-AI works in production. In legal I told you to watch the M&A instead of the messaging. In accounting, the M&A comes with an AI lab attached.

Here is how I think it ends.

The consolidators win the middle. In legal I called the squeezed middle acquisition inventory and could not name the buyer. Here the buyers are named, funded, and holding production data. The retirement cliff supplies the sellers; the renewing revenue makes the maths work; the AI makes the margins.

The AI-native firms win the new floor. The millions of small businesses that never had a proper finance function... too small for the old cost structure... become servable for the first time. That market goes to the natives, not to anyone's roll-up.

The Big Four hold the ceiling by becoming operators. Audit and complex advisory stay theirs, and managed services quietly turns them into the largest AI-native service providers on earth without anyone using the phrase.

The tools face the squeeze, again. Priced as category winners, caught between incumbent software above and buyers who build their own AI inside acquired firms below. Same hardest decade, different vertical.

And one dated prediction: within twelve months, this exact template... permanent capital plus embedded lab... gets pointed at legal and at insurance brokerage. If I am wrong, it will be obvious by next summer.

If you are a lawyer who read last week's edition: the buyer your market lacks now has a proven playbook and fresh capital, and Thrive's own second platform is already rolling up IT services firms. The middle of every professional market should assume the chequebook is coming.

And if you read my post last week on Microsoft, Amazon, OpenAI and Anthropic committing $9 billion to move into services... notice that this edition is the same story from the other end. The model lab is no longer just deploying engineers into businesses. In accounting, it owns part of the buyer. The ladder from selling tools to owning the work is not a forecast anymore. It has a portfolio.

That is the pattern this newsletter exists to track... not one market's drama, but the same war, fought with different armies, wherever the work is worth more than the software.

One more thing. This was meant to be a biweekly letter. The category is moving too fast for that, so the first four editions will be weekly. We settle back once the news does. Subscribe, and if you think the rankings are wrong, reply and tell me why. See you next Tuesday.