The Bench Founder Built the Category. Now He's Killing It.

01 /The Concept

Ian Crosby built Bench Accounting from a Vancouver startup into the largest bookkeeping service for small businesses in North America over twelve years, serving tens of thousands of customers. Then, in late 2024, Bench stopped accepting payments, laid off its staff, and shut down operations before being acquired within days to prevent a total collapse. Crosby founded Synthetic the following year. Synthetic employs no bookkeepers.

Synthetic connects to a company's bank accounts, payroll system, billing platform, and email inbox. It asks a set of clarifying questions to understand the specific accounting treatment of the business – recurring revenue versus project-based, capitalized software expenses, intercompany transactions – and then delivers a clean set of accrual-basis books ready for a tax preparer. No human accountant reviews the output. The service starts at $49 per month, roughly a quarter of what Pilot charges at its lowest tier. Pilot, the YC-backed bookkeeping service that raised approximately $140 million targeting the same customer segment, starts at $499 per month.

Synthetic's target is narrow by design: only software, SaaS, and AI companies. Those businesses have structurally cleaner financial data than most small businesses – MRR from subscription platforms, payroll from a single vendor, cloud spend from AWS or GCP – which makes the integrations reliable enough for AI to produce accurate accrual-basis books without human review. Khosla Ventures led the $10 million seed. Tobi Lütke, Kaz Nejatian, and Michael Tannenbaum participated.

02 /Why It Matters

What Crosby understands about the economics of bookkeeping that most observers miss is where the service actually fails. Bench did not collapse because small businesses stopped needing bookkeeping. It collapsed because the model of augmenting human bookkeepers with software creates a cost structure that scales with revenue: more customers require more bookkeepers, and bookkeeper wages were rising faster than the pricing tolerance of the small business market. Bench's growth compelled it to hire; its hiring compelled it to raise prices; its price increases compelled its customers to churn to cheaper alternatives. The loop was structural.

Synthetic's model, if the AI accuracy holds, breaks the loop. Revenue grows without adding headcount, and the marginal cost of bookkeeping each additional company should decline over time as the models improve on more data from similar business structures.

The narrow focus on SaaS companies is a technical constraint before it is a market segmentation decision. Accrual-basis bookkeeping for a restaurant involves cash tips, variable food costs, supplier invoices with irregular timing, and payroll across a high-turnover workforce – judgment calls at the transaction level that AI cannot currently handle reliably. A SaaS company with subscription billing, payroll from Rippling, and a handful of recurring vendor relationships has the kind of clean, structured financial data that makes autonomous bookkeeping tractable. The narrow market is the market where the AI quality can actually meet the accuracy threshold that makes the product genuinely useful.

The Pilot comparison is the most consequential competitive dynamic. Pilot raised $140 million building a large team of accountants who use proprietary software to complete work faster than a traditional firm. If Synthetic can deliver comparable accuracy at $49, Pilot's model has a structural problem in the segment that constitutes the majority of its customer base.

03 /Opportunities

Accuracy is the product gate. Bookkeeping that is 90% accurate costs more to fix at tax time than the price differential saves across the year. The early cohort of SaaS companies adopting Synthetic will generate the ground truth data to assess whether the AI accuracy meets professional-review standards. That cohort is also the customer segment most likely to tolerate early-stage quality variance – they chose software over services for their CRM, payroll, and HR; they will do the same for bookkeeping if the quality threshold is met.

The integrations that make bookkeeping accurate also make adjacent services tractable. Automated tax filing, cash flow forecasting, and vendor payment management each require the same bank, payroll, and billing data connections already established. Pilot built those services with large teams at correspondingly high cost. Building them on top of autonomous bookkeeping infrastructure creates a faster development path and a structurally lower cost structure than any human-augmented competitor can match. The SaaS focus is where data quality is highest and adoption tolerance is greatest; the expansion into adjacent financial services follows the same data integrations already in place.