Accrual announced on 2 September that it is acquiring Puzzle’s accounting-firm technology and business — the AI-native ledger, the close products, and Sasha Orloff himself. Terms were not disclosed. Puzzle keeps its brand and its small-business customers; what it gives up is the right to sell a firm-facing ledger to CPA firms.
Every recap of this deal you’ll read this week is written for accountants. This one is for the founder with a board meeting on Thursday and a roadmap built on the assumption that firms will eventually move their clients somewhere better.
Our read: Puzzle didn’t lose on product. It lost on the unit of sale.
TL;DR
- The ledger wasn't the problem. Puzzle built a credible AI-native ledger. Adopting it required a firm to migrate clients, and no amount of product quality makes that a small decision.
- The base doesn't move. Xero held monthly churn at 1.14% in FY26 while lifting ARPU 23%. Intuit grew QuickBooks Online Accounting revenue 23%. Firms absorbed price rises rather than switch.
- Xero already proved what does move a base. WorkflowMax in 2012 became Xero Practice Manager; Xero Tax shipped free to partners in 2015 and synced from the client's Xero file. Firms migrated clients because their own workflow got faster, not because the ledger was better.
- Firms bought labour, not ledgers. Accrual went at tax prep and had Top 100 firms inside seven months. Basis raised at $1.15bn for agents that install on the ledger a firm already runs.
- The stock market is telling a different story to the customers. Intuit shares fell roughly 38% across 2026 on AI-disruption fear while its accountant-channel revenue compounded. Don't confuse a multiple with a moat.
- Price against the migration. If your product only works once clients move in bulk, your real competitor is inertia, and inertia is winning.
What actually happened
The facts, as of 4 September 2026, and only the ones that are public.
Accrual launched in February 2026 with $75m led by General Catalyst — which, per Bloomberg’s reporting at the time, identified accounting as a category to attack, recruited former Brex CTO Cosmin Nicolaescu, and built the company internally. It went to market against tax preparation and review, and by the time of this announcement counted Armanino, Aprio, H&R Block, Creative Planning, BMSS and Stephano Slack among its firms.
Puzzle was founded in 2019, has raised roughly $65m, and serves 7,000-plus startups, small businesses and firms. Accrual is acquiring the firm-facing technology and business. Orloff and team members join Accrual; Ozgur Uzuner takes over the remaining small-business company as CEO with Jason Mitchell as CTO. Accrual plans broader availability of the combined client accounting services capability by the end of 2026.
One detail the trade coverage skipped: General Catalyst was on both sides. It backed Puzzle and it built Accrual. This is a fund consolidating its own accounting bet into the vehicle it controls, and choosing which thesis to keep.
The thesis it kept was not “replace the ledger.”
The two numbers that explain the whole deal
Xero’s FY26 results, to 31 March 2026: operating revenue up 31% to NZ$2,753.1m, 4.92 million subscribers (up 11%), ARPU up 23% to NZ$55.44, annualised MRR up 37%, and monthly churn of 1.14%.
Read those together rather than separately. Xero spent that year raising prices, retiring App Store subscription billing, and metering its API by connections and data egress — changes we walked through in what Xero and Intuit absorbed in 2026. It took revenue per customer up 23% and lost almost nobody.
Intuit’s fiscal 2026, reported on 1 September: revenue $21.45bn, up 14%. QuickBooks Online Accounting revenue up 23% for the year and 20% in Q4, driven by “higher effective prices, customer growth and mix shift”. And the line that should stop any ledger-replacement founder cold: accountants drove 25% of Intuit’s new Enterprise Suite contracts.
Accountants aren’t a switching force at the incumbents. They’re a distribution channel for them.
If a materially better ledger existed and firms were willing to move for one, you would see it in churn. Instead you see a base that absorbed price rises without flinching. That’s the empirical answer to the question Puzzle’s roadmap was betting on, and it was available before the deal.
A better ledger is not a problem a firm has
This is the part we’d argue with a founder over a whiteboard.
A firm’s ledger is its deepest control point — and critically, it isn’t the vendor’s control point over the firm, it’s the client’s history holding the firm in place. Years of coded transactions, a chart of accounts that reflects how that business actually works, bank feeds that took three phone calls to establish, an app stack hanging off the API, and staff who can navigate it half-asleep at 11pm in deadline season.
To adopt a replacement ledger, a firm doesn’t make a software decision. It runs a project, per client, and gets nothing back until it has run that project across a meaningful share of the book. The economics only work in bulk, and bulk is exactly what a partner group will not authorise on the strength of better categorisation.
That’s the gap Puzzle’s firm-facing product could never close. It wasn’t a marketing gap. You cannot message your way past a migration.
Notice too what firms told the AICPA they were worried about. The 2026 CPA Firm Top Issues Survey — 629 respondents, fielded April to May 2026 — put managing change from technology and AI as the leading five-year issue, ranked first among the largest firms, with staffing and retention still dominating the near term. A firm naming change management as its number one risk is a firm telling you, in writing, that it will not be migrating anything in bulk.
Xero already ran this experiment, and won it
The most useful precedent for this deal is fourteen years old, and it’s the one that built Xero’s Australian business.
In May 2012 Xero acquired WorkflowMax, a job and practice management tool. Its technology became Xero Practice Manager. In 2015 Xero added Xero Tax and gave it away free inside the partner programme, going directly at Sage HandiSoft, MYOB and Reckon — who had been selling firms comprehensive tax packages with very little innovation in them. By 2019 Xero was reporting more than 5,000 firms across Australia and New Zealand filing over six million returns through it.
None of that was ledger work. It was the firm’s own workflow: jobs, capacity, lodgement, billing.
And here’s the detail that turned it from a useful add-on into a land grab. Xero Tax synced automatically from the client’s Xero file. The practice tools were free, and they worked best when the client sat on Xero’s ledger. So firms didn’t migrate clients because Xero’s ledger was better than MYOB’s. They migrated because their own workflow got faster, and the ledger came along behind it.
A Xero advisor quoted at the time of the WorkflowMax deal described the mechanism more plainly than any strategy deck since: running his practice and his clients in one place saved time, improved efficiency and “allowed me to take on more clients, doubling my business”.
That is what a problem big enough to move clients in bulk actually looks like. The problem was the firm’s capacity. The bulk migration was the side effect.
Someone is running that playbook again right now
Which is why the company worth watching in this cycle isn’t an AI ledger sold as an AI ledger.
COUNT is building general ledger, practice management, a monthly close wizard, client portal and time-and-billing as a single platform, positions itself as “the operating system for accounting firms”, and has a partner programme flagged as coming. That is the 2012 Xero shape with an AI-native ledger underneath it — and it is a far better-formed bet than a ledger sold on its own merits, because the ledger isn’t the pitch. The firm running better is the pitch. The ledger is just where the value accrues once it works.
Whether COUNT executes is a separate question and it’s early days. But the shape is right, and the shape is the thing Puzzle’s firm-facing product never had. Puzzle asked firms to move clients in order to get the benefit. Xero gave firms the benefit and collected the clients afterwards.
This is also why the partner programme is a distribution decision rather than a marketing one. Free practice tooling inside a partner tier isn’t generosity — it’s the cheapest client acquisition in this industry, paid for by the ledger seats that follow.
What firms did buy in the same twelve months
Two comparisons make the point better than any argument.
Accrual. Launched February 2026. Attacked tax preparation and review — the single largest pile of billable, unhireable labour in a US firm. Top 100 logos inside seven months. It never asked a firm to move a client anywhere.
Basis. Raised $100m at a $1.15bn valuation in February 2026, led by Accel with GV and Khosla participating. Its agents run across CAS, tax, audit and advisory, and they connect to QuickBooks Online, Xero, NetSuite, SAP and Plaid bank feeds — typically live in under 48 hours. It reports working with 30% of the top 25 firms.
Same month, same category, same “AI-native” language. One raised at $1.15bn for agents that sit on the ledger a firm already runs. The other spent seven years building the ledger itself and sold that business for undisclosed terms.
The difference isn’t model quality. It’s that one of them starts paying on Tuesday and the other starts paying after a two-year migration programme. We’ve written before about why time to value decides adoption in this channel, and this is the most expensive demonstration of it we’ve seen.
The market’s read and the customer’s read have come apart
Worth flagging, because it will show up in your next fundraise.
Intuit shares fell roughly 38% across 2026 as of early September, on a broad AI-disruption trade rather than anything in the results — JPMorgan cut to Neutral and took its target from $605 to $331. Meanwhile the accountant-channel revenue inside that same company grew 23%.
So public markets are pricing incumbent accounting software as if the base is about to move, and the base is not moving. Both things are true at once and they point in opposite directions for a founder.
The investor-facing story — “the ledger is up for grabs, AI resets the incumbency” — is currently easier to raise on than to sell on. Puzzle raised well on it. It could not sell on it. If your deck leans on that narrative, be honest with yourself about which of those two you’re optimising for, because they have started to diverge.
Six things to take into your roadmap
1. Price your product against the migration, not against the incumbent’s feature list. The question is never “is ours better than QuickBooks”. It’s “is ours better by more than the cost of moving 400 clients”. Almost nothing clears that. Products that require no move clear a bar an order of magnitude lower.
2. Work out who inside the firm can say yes. A ledger swap needs the partner group, the CAS lead, IT and every client. A tool that runs on the existing stack needs one champion with a budget. That difference is the whole sales cycle — it’s the core of how the firm buying committee actually behaves.
3. Sell hours, not accuracy. Capacity is the pain that moves firms in bulk, because it’s the one they cannot hire their way out of. “More accurate books” is a quality claim, and firms already believe their books are fine.
4. If you want the ledger, sell the workflow first. This is the Xero manoeuvre and it still works. Own the job list, the close, the lodgement or the billing; make your tooling materially better when the client sits on your data; then let the firm move clients at its own pace for its own reasons. You get the same migration Puzzle needed — you just don’t have to ask for it up front.
5. Assume “AI-native” is a feature, not a wedge. Xero shipped XeroForce and a pre-built month-end agent in 2026; Intuit is shipping agents across the suite. Anything whose only differentiation is that the AI is newer gets absorbed. The durable positions are the ones we mapped in where to build after the platform announcements — vertical depth, the exception layer, jurisdiction-deep compliance, and systems of action.
6. Be multi-ledger earlier than feels necessary. Basis’s install list is a competitive weapon, not an integration backlog. A firm’s real client list is spread across Xero, QuickBooks, MYOB and Sage, and a product that only describes part of it only ever gets part of the firm.
The verdict
We don’t know whether this was a good sale. Nobody outside it does — the terms aren’t public, and anyone telling you it was a fire sale or a triumph is guessing. Puzzle raised around $65m, sold one business line rather than the company, kept its brand and its SMB customers, and put its founder inside a well-funded acquirer with Top 100 distribution. That’s a real outcome, and it’s a better one than most infrastructure-replacement plays in this channel get.
What we would say plainly is this: the outcome was legible in the churn data long before it was legible in the press release. A category where the incumbent lifts price 23% and loses 1.14% of subscribers a month is a category telling founders that replacement is not the game.
And the alternative was never a secret. It’s sitting in Xero’s own history — buy the practice management tool, give the firm its hours back, make the tooling best when the client sits on your data, and let the ledger follow. Xero didn’t out-feature MYOB and Reckon. It made itself the way an Australian firm ran its practice, and the client files came with it.
So if you’re building something that only works once accountants move clients in bulk, that’s the assumption to stress-test in your next planning session — not the model, not the UI, the assumption. Then go looking for the workflow you could own instead, because that’s the door the ledger opens from.
That stress test is most of what we do when we’re brought in on go-to-market strategy for accounting software: find the version of the product a firm can say yes to on a Tuesday, and work out what it earns you on the Tuesday after that.
Sources: Accrual and Puzzle announcements via CPA Practice Advisor and Accounting Today (2 September 2026); Xero FY26 annual results (14 May 2026); Intuit fiscal 2026 fourth-quarter results (1 September 2026); Basis Series B announcement (24 February 2026); AICPA PCPS CPA Firm Top Issues Survey 2026; Xero’s WorkflowMax acquisition as reported May 2012; COUNT product positioning as published September 2026.
