Botkeeper Shut Down Overnight. What Firms Should Demand From Any Bookkeeping Vendor Now

August 11, 2026

---

In early February 2026, Botkeeper — roughly eleven years old, with roughly $90 million raised — shut down abruptly. Firms that had built their workflow around it woke up to a vendor that no longer existed. Weeks later, Xendoo acquired the Infinite platform, which resolved the corporate question but not the one that matters to a practice owner: what do you do the morning your bookkeeping vendor disappears, and what should you have asked before you ever signed?

This is not a schadenfreude post. Botkeeper employed real people and served real firms for over a decade, and eleven years is longer than most software companies survive. This is a vendor-risk post, and the checklist below applies to every vendor in this category — including us. Apply it ruthlessly.

What actually happened to Botkeeper?

The short version: an abrupt shutdown in early February 2026, after about eleven years of operation and about $90 million in funding, followed weeks later by Xendoo acquiring the Infinite platform. The abruptness is the part worth studying. This was not a long wind-down with a migration window and a deprecation calendar. Firms with client books in motion — month-end closes half done, reconciliations in progress — had to figure out continuity in real time.

For the firms affected, the acquisition of the platform weeks later was better than nothing. But "weeks later" is an eternity when you have clients whose books need to close this month and a team whose entire process ran through a tool that just stopped answering.

How does a company with $90 million raised shut down overnight?

Funding is not the same as durability. The number that explains the fragility better than the funding total: reportedly 30 to 40 percent of Botkeeper's revenue was concentrated in roughly ten enterprise clients.

Sit with that structure for a moment. When a third or more of revenue sits with ten accounts, the departure of two or three of them is not a bad quarter — it is an existential event. And the hundreds of smaller firms on the platform, the ones who were individually reliable and collectively loyal, had no visibility into that concentration and no vote when it mattered.

The lesson is not "avoid venture-backed vendors." Plenty of bootstrapped companies die too, usually quieter deaths. The lesson is that a vendor's survival depends on facts you cannot see from the pricing page: who their revenue actually comes from, and what happens to the whole platform if the biggest slice walks. Which is why the questions below are worth asking out loud, even when they feel awkward.

What is the difference between losing a tool and losing your books?

Everything, and it comes down to one architectural question: where does the ledger live?

If a vendor is an overlay — it reads from and writes to a ledger you control, and that ledger remains intact without it — then a shutdown costs you a workflow. Painful, disruptive, recoverable. You lose speed. You do not lose the books.

If a vendor is the system of record — the ledger itself lives inside their platform — then a shutdown costs you the books, or at minimum puts them behind a door someone else holds the key to. This is not a hypothetical distinction in this market. Digits, one of the AI-native players courting firms right now, states it plainly in its own positioning: after cutover, "Digits becomes your system of record," with the old QuickBooks Online connection archived. That may be a fine trade for some firms. But it should be a trade you make with your eyes open, because it changes what a vendor failure costs you from inconvenient to catastrophic.

Ask every vendor which one they are. Then ask what a full export looks like — not in the sales deck, in practice.

What should you demand from any bookkeeping vendor now?

Five things. In writing, before you commit a single client.

1. Data export, on demand, in a usable format. Not "contact support for an export." Not a proprietary archive that only their platform can read. You should be able to pull complete client data yourself, today, without asking permission — because on the day you actually need it, there may be nobody left to ask.

2. A straight answer on who owns the ledger. Overlay or system of record. If the answer is system of record, ask exactly what a reverse migration looks like and how long it takes. If the answer is vague, treat vague as a no.

3. A conversation about revenue concentration. This one feels intrusive, and most vendors will not hand you a customer list. Fine. Ask the shape of the question instead: what fraction of your revenue comes from your ten largest customers? A vendor with a broad base can answer comfortably. A vendor who bristles is telling you something. Botkeeper's firms never got to ask.

4. Verifiable work, not a percentage to trust. Some vendors in this category now sell automation as a statistic — Digits, for instance, offers outcome-based pricing where you pay only when at least 95 percent of transactions are zero-touch. Notice what that pricing model asks of you: trust that the untouched 95 percent was right. Practitioners already know how that story tends to end, and they say so in the same recurring phrases — "I still have to check everything anyway," "it looked right until we reconciled," "the books balance but the reports are wrong." Demand the opposite structure: work you can inspect line by line, with the evidence for each decision shown, so checking is fast instead of a leap of faith.

5. A parallel run before commitment. Run the vendor alongside your existing process on real clients for a full month and compare outputs before you move anything. Any vendor confident in their product should welcome this. Notably, almost nobody in the firm-focused AI bookkeeping market publishes one — or publishes much of anything. Puzzle and Truewind list no public pricing at all; Booke publishes a per-business price but keeps firm pricing behind a sales call; Puzzle claims SOC 2 compliance without publishing whether it is Type I or Type II. None of this makes them bad products. It does mean the burden of the awkward questions falls entirely on you.

How do you actually ask a vendor these questions?

By email, in writing, before the demo charm offensive begins. Five questions: Can I export all client data myself, on demand, in a standard format? Are you my system of record or an overlay on a ledger I control? What share of your revenue sits with your ten largest customers? Can I inspect and verify individual transactions, or am I trusting an accuracy percentage? Will you support a parallel run on my real clients before I commit?

Then judge the answers on two axes: substance, and whether they were willing to put the answer in writing at all. A published price, a published pilot, a published architecture — these are not marketing niceties. They are a vendor pre-committing to answers before you asked, which is exactly the behavior you want from someone who will hold a piece of your practice. "Talk to us" is a legitimate sales motion and also, sometimes, a way of never being on the record.

Where Nalo fits, stated plainly

We built our Botkeeper-alternative page to answer this exact checklist about ourselves: your ledger stays yours, every categorization ships with the evidence behind it, nothing posts without confirmation, and the parallel-run evaluation month is published, free, and open to any firm — including firms that ultimately choose someone else. Hold us to the same standard you now hold everyone. That is the point of the checklist.

Botkeeper's shutdown was a bad morning for a lot of good firms. The only useful response is to make sure the next abrupt shutdown — whoever it happens to — costs you a workflow, and never the books.