AI bookkeeping has a familiar pitch and an unfamiliar salesman. Tabby, profiled by TechCrunch on 21 September 2026, was built by Ahad Ali, a New York CPA who previously ran a practice with a 20-person team filing more than 2,000 tax returns a year. The headline framing is that he is automating away the job he used to do, and that AI bookkeeping is the thing that finally does it.
Fourteen months in, the company reports 5,500 small businesses on the platform, roughly $100,000 in annual recurring revenue, a team of seven, and a $1 million pre-seed round in progress. The AI bookkeeping platform imports live bank data through Plaid and builds a running profit-and-loss view rather than a month-end reconciliation.
That is the story as told. Tabby’s own website tells a noticeably different one, and the difference is checkable in about ten minutes with a browser and a calculator. This article sets out what the AI bookkeeping product actually does, what the reported numbers divide into, and why the company’s own pages describe a business that sells to accountants rather than one that replaces them.
Table of contents
- What Tabby’s AI Bookkeeping Actually Does
- The AI Bookkeeping Numbers, and What They Divide Into
- The Page That Contradicts the Headline
- Two Numbers on Tabby’s Own Page That Sit Awkwardly
- What AI Bookkeeping Replaces, and What It Does Not
- How the AI Bookkeeping Plans Are Priced
- What This Means for Firms Here
- Frequently Asked Questions About Tabby and AI Bookkeeping
- References
What Tabby's AI Bookkeeping Actually Does
The mechanism of any AI bookkeeping tool matters, because it determines exactly what can and cannot be automated out of a practice.
The data pipeline
Tabby connects to bank and card accounts through Plaid and pulls transactions on a schedule. The AI bookkeeping layer categorises each transaction, learns rules from corrections, and keeps a profit-and-loss statement current rather than producing one after the quarter closes. The company markets the output as tax-ready reporting.
What the vendor says it is for
Tabby’s own machine-readable site summary is unusually specific about the target customer. The product is “built for solopreneurs, gig workers, 1099 contractors, Uber drivers, and interdependent realtors like you to calculate your taxes without hiring an accountant.” That is a description of people who do not currently have an accountant, not of people leaving one.
Features on the roadmap
A natural language interface called Tabby Talk is in development, which would let an owner ask questions of their own ledger rather than reading a dashboard. The platform also advertises one-click migration from QuickBooks, which is the incumbent it names directly, and data export in CSV, QBO and PDF at any time.
The founder’s stated thesis
Ali’s argument is about the shape of demand rather than the technology. Small operators do not want a better ledger interface, he told TechCrunch: “They need something that just does it for them.” On that reading AI bookkeeping becomes a utility bill — a standing monthly charge on the order of internet access or payroll, rather than a professional relationship.
Where the marketing actually points
The clearest signal of intended customer is not the homepage but the blog. Tabby’s published posts are overwhelmingly about residential property transactions: seller closing costs by state, how to read a seller’s net sheet, how to present one to win a listing. The AI bookkeeping product has a named vertical, and it is estate agents rather than accountancy practices.
The competitive frame
Tabby names QuickBooks as the incumbent and offers one-click migration from it. The venture-funded comparison is starker: Sequoia-backed Rillet raised $100 million for accounting automation aimed at larger finance teams. Against that, a $1 million pre-seed puts Tabby in a different weight class entirely, chasing a segment the well-funded entrants find too small to serve.
The AI Bookkeeping Numbers, and What They Divide Into
Two reported AI bookkeeping figures sit next to each other and are worth dividing rather than reading separately.
Revenue per user
Roughly $100,000 in annual recurring revenue spread across 5,500 AI bookkeeping accounts is $18.18 per business per year, or $1.52 a month. No plan Tabby sells costs anything near that. The cheapest paid tier is $19 a month billed yearly, which is $228 a year.
What that implies about the paid base
Dividing $100,000 by $228 gives a ceiling of about 438 paying accounts. Against 5,500 users that is 8.0%. If the average payer is on the Standard monthly plan at $29, the figure is 287 accounts, or 5.2%. On Professional monthly at $39 it falls to 214, or 3.9%. The AI bookkeeping product is overwhelmingly running on its free tier.
The user count disagrees with itself
TechCrunch reports 5,500 small businesses on the AI bookkeeping platform. Tabby’s own pricing page, read on 22 September 2026, says “4,500+ operators stopped dreading their books” and repeats the figure in its testimonial block. That is a thousand-user gap between the pitch and the shopfront, or about 18% of the larger number.
Why none of this is damning on its own
A free tier carrying most of the base is the normal shape for a fourteen-month-old AI bookkeeping product raising a pre-seed round, and the two user counts may simply have been written months apart. The point is that $100,000 of ARR and 5,500 users describe a product in distribution, not a product displacing an industry. The AI bookkeeping revenue is roughly what one accountant bills in a quarter.
The Page That Contradicts the Headline
If the thesis is that AI bookkeeping makes accountants obsolete, the company’s own site navigation is the first problem with it.
Tabby sells a product to accounting firms
Tabby’s main menu carries an “Accountants” item alongside the AI bookkeeping product pages. Behind it sits a separate product, “Bookkeeping Software for Accountants, CPAs and Bookkeepers”, strapped “AI bookkeeping, built by a CPA who ran a firm just like yours.” The page advertises 100+ accounting firms already on it, a free firm licence, and pricing at $15 per client per month.
The pitch is augmentation, in the founder’s own words
The page’s headline argument is addressed firm-owner to firm-owner: “You didn’t build a firm to be a bookkeeping shop. You built it for the work that compounds — tax strategy, advisory, judgment your client can’t get from software. Tabby is built to give your team that work back.” The section heading is “From firm-as-bookkeeper to firm-as-advisor.”
Accountants are an AI bookkeeping sales channel, not a target
The same page offers partner pricing and a referral commission, magic links so clients can connect banks without an account, bulk categorisation, and team dashboards for assigning clients to staff. It claims the AI bookkeeping layer automates 80% of the tedious work, cuts client back-and-forth by 70%, and saves up to 10 hours per client.
| Claim | The headline framing | What Tabby’s own pages say |
|---|---|---|
| Who the customer is | Businesses leaving their accountant | Solopreneurs and 1099 workers “without hiring an accountant” |
| Relationship to firms | Displacement | A firm product, a free firm licence and a referral commission |
| Firms already using it | Not stated | 100+ accounting firms |
| What the work becomes | Obsolete | “From firm-as-bookkeeper to firm-as-advisor” |
| Share of work automated | Implied total | 80% of “the tedious work” |
| Named competitor | The accounting profession | QuickBooks, with one-click migration |
What the firm price implies about the channel
At $15 per client per month, a firm seat is $180 a client a year. If the entire $100,000 of ARR came from that channel it would represent about 556 client seats across 100+ firms — roughly 5.6 clients each. Either the firm channel is barely switched on, or the revenue is mostly direct. Both readings put the AI bookkeeping business well before the point of displacing anything.
Two Numbers on Tabby's Own Page That Sit Awkwardly
The AI bookkeeping pricing page carries three headline statistics, and two of them pull against each other.
The claims as written
Directly beneath the testimonials, Tabby states “$15M+ in deductions flagged by Tabby”, “$5,000 average tax savings per user, per year”, and “10+ hrs saved per user, every month”. The same page gives the user base as 4,500+.
The arithmetic
A deduction reduces taxable income; the tax saved is the deduction multiplied by a marginal rate. Multiplying $5,000 by 4,500 users gives $22.5 million of tax saved — from $15 million of deductions found. At a 30% blended marginal rate, $15 million of deductions saves roughly $4.5 million in total, or about $1,000 per user. Reaching $5,000 each would need around $75 million of deductions.
The charitable readings
There are two available, and both are plausible. The $5,000 may be an average across paying users only, which the earlier division suggests is a few hundred accounts rather than 4,500 — that would reconcile the figures without much strain. Or “tax savings” may be loose phrasing for deductions surfaced, in which case the two numbers are the same quantity described twice.
Why it still matters
Neither reading is signposted, and a prospective buyer doing the multiplication gets a number that cannot be right. For an AI bookkeeping product whose entire value proposition is arithmetic done correctly on your behalf, the marketing arithmetic is the first thing a sceptical accountant will check. It is a free credibility win the company has left on the table.
What AI Bookkeeping Replaces, and What It Does Not
The honest AI bookkeeping boundary is narrower than either the headline or the vendor implies, and it is well established in practice.
The part that genuinely automates
Transaction categorisation, rule learning, bank reconciliation, receipt matching and report generation are pattern work on structured data with a correction loop, and they are what AI bookkeeping is genuinely good at. This is the strongest possible case for automation, and Tabby’s claimed 80% figure for “the tedious work” is not obviously overstated. Recurring transactions in a small business are highly repetitive.
The part that does not
Judgement calls do not automate, and no AI bookkeeping accuracy rate changes that: whether an expense is genuinely allowable, how to treat a mixed-use asset, what a director’s loan account should look like, whether a filing position will survive scrutiny, and what to do when the answer is arguable. Tabby’s own firm page names exactly these — “tax strategy, advisory, judgment” — as the work it hands back.
The compliance boundary AI bookkeeping cannot cross
An AI bookkeeping tool can produce a set of figures. It cannot sign them. Statutory accounts, audit, and representation to a tax authority all require an identified responsible person, and that requirement is legal rather than technical. No categorisation accuracy rate changes it.
Where the pressure actually lands
The squeeze is on the low-margin compliance base of a practice: the monthly bookkeeping retainer for a sole trader with one bank account. That work has been under price pressure for a decade, and AI bookkeeping accelerates it rather than starting it. The right response for a firm is the one Tabby’s own page argues for, which is to stop selling hours of data entry.
How the AI Bookkeeping Plans Are Priced
The AI bookkeeping plan structure is worth reading closely, because the expense ceilings tell you exactly who each tier is built for.
The three tiers
Free Forever is $0 a month and caps annual expenses at $15,000 with one connected account. Standard is $29 a month, or $19 billed yearly, capping annual expenses at $200,000 with three accounts and a single user. Professional is $39 a month, or $29 billed yearly, with unlimited expenses, up to five users and unlimited accounts.
| Feature | Free Forever | Standard | Professional |
|---|---|---|---|
| Monthly price | $0 | $29, or $19 yearly | $39, or $29 yearly |
| Annual expense ceiling | $15,000 | $200,000 | Unlimited |
| Connected accounts | 1 | 3 | Unlimited |
| Users | 1 | 1 | Up to 5 |
| Sync frequency | Daily | Daily | Real-time |
| Reporting | P&L only | P&L and tax-ready | Advanced financial reports |
The jump between the tiers
The AI bookkeeping expense ceiling rises from $15,000 to $200,000 between Free and Standard, a factor of 13.3. That gap is where the conversion happens: a genuine sole trader with a van and a phone contract will stay free indefinitely, and anyone with real cost of sales crosses it in the first year. The free tier is a funnel, not a product.
Value per hour claimed
Taking the advertised 10+ hours of AI bookkeeping saved a month at face value gives 120 hours a year. Against the $348 annual cost of Standard billed monthly, that is $2.90 per hour saved. Against $228 on the yearly plan it is $1.90. Those are strong numbers if the hours are real, which is a question only a user’s own timesheet answers.
How the trial converts
The mechanics are worth knowing before signing up. The trial runs 14 days and the card is charged on day 15 rather than at the end of a grace period. A 30-day money-back guarantee sits behind that, and cancellation is described as two clicks in billing settings with data export in CSV, QBO and PDF available before, during or after. That is a fair structure, and the export promise is the important part — an AI bookkeeping tool holding your ledger hostage would be the real lock-in risk.
What a UK reader should adjust for
Tabby is a US AI bookkeeping product and its language is American throughout: 1099 contractors, IRS-ready reports, Plaid connections to US banks. The equivalents here are Self Assessment, Making Tax Digital and open banking, and the compliance detail does not transfer. Treat the pricing as a signal of where the market is heading rather than as a quote.
What This Means for Firms Here
The strategic reading of AI bookkeeping matters more than the product review, and the direction is not new.
The work that is genuinely at risk
If a practice’s revenue depends on charging for transaction coding, AI bookkeeping is the direction of travel and it will not reverse. Firms that already made the move to advisory will find this story unremarkable; the ones that did not now have a named, funded, seven-person example of the pressure. Our guide to Microsoft Copilot for accountants covers the same shift from inside the incumbent toolset.
The integration question
Practices adopting any AI bookkeeping tool inherit a data-flow problem before they gain a time saving: bank feeds, practice management, tax software and document storage all have to agree with one another. That is ordinary business process automation work, and the cost of getting it wrong is usually larger than the licence fee. Budget for the plumbing, not just the subscription.
The security question nobody asks first
Handing a third-party platform read access to every client bank account is a material risk decision, and it is the one least likely to appear in a demo. Firms subject to Cyber Essentials requirements should treat an AI bookkeeping vendor as they would any other data processor, with the same due diligence and the same contractual position on breach.
The fair verdict on Tabby
Tabby is a fourteen-month-old AI bookkeeping company with a seven-person team, $100,000 of ARR, a free tier carrying most of its users, and a second product line sold to the profession it is described as replacing. That is a credible early-stage business and an implausible extinction event for the profession.
The interesting thing about it is not the threat; it is that the founder is hedging, and the hedge is on his own website. A CPA building AI bookkeeping software has better information than most about which parts of a practice are safe, and he has chosen to sell to firms as well as around them. That is the most informative fact in the story, and it is the one the headline leaves out.
Frequently Asked Questions About Tabby and AI Bookkeeping
Does Tabby’s AI bookkeeping replace an accountant?
Not on the company’s own account. Its firm-facing product is pitched at moving practices from bookkeeping to advisory, and its consumer positioning targets self-employed people who do not currently use an accountant at all.
How much does Tabby cost?
There is a Free Forever tier capped at $15,000 of annual expenses, Standard at $29 a month or $19 billed yearly, and Professional at $39 a month or $29 billed yearly. Firms pay $15 per client per month for the AI bookkeeping firm product, with a free firm licence.
How many customers does Tabby have?
TechCrunch reports 5,500 small businesses. Tabby’s own pricing page says 4,500+ operators. The company also states 100+ accounting firms on its firm product.
Is $100,000 in ARR a lot for this stage?
It is modest and appropriate for a pre-seed company fourteen months old. Divided across the reported AI bookkeeping user base it is about $18 per user per year, which indicates most users sit on the free tier rather than that the pricing is broken.
What does AI bookkeeping automate well?
Categorisation, rule learning, reconciliation, receipt matching and report generation — repetitive pattern work on structured data with a correction loop. Accuracy improves with volume, which is why every incumbent is adding the same AI bookkeeping features.
What can it not do?
AI bookkeeping cannot make or defend a judgement call, and it cannot sign anything. Filing positions, allowability questions, statutory accounts and representation to a tax authority all require an identified responsible person.
Should a UK firm adopt a US AI bookkeeping tool?
Only with care. The AI bookkeeping compliance layer is built for IRS reporting and US bank connections, so the outputs will not map cleanly onto Self Assessment or Making Tax Digital obligations without additional work.
