Bookkeeping vs Accounting Software: Why AI Is Splitting the Bundle

Windsurf, one of the fastest-growing AI coding companies, runs its entire finance operation with two people. Postscript, a SaaS business doing more than $100M in ARR, closes its books in three days. Both run on Rillet, an AI-native ERP that has raised more than $100 million in funding inside its first year. Look past the funding to the two-person finance team.

For thirty years, bookkeeping and accounting software meant the same thing. QuickBooks. Xero. Sage. NetSuite. One product, one login, one ledger. That bundle is coming apart, and the unbundling matters for any SaaS, vertical SaaS, or embedded finance product that touches financial data.

Bookkeeping and accounting software: always two jobs

Bookkeeping is the data layer. Recording transactions, categorizing expenses, matching payments to invoices, reconciling bank feeds. Mechanical work at high volume, where errors break everything downstream.

Accounting is the interpretation layer. Producing financial statements, running the month-end close, applying revenue recognition rules, generating reports for boards and lenders. Judgment work that depends on the bookkeeping being clean. If you build software that talks to accounting systems, the data model of every accounting API is shaped by these rules.

QuickBooks bundled both because in 1992 there was no practical way to separate them. The bookkeeping data lived inside the product, so the reporting had to live there too. Xero did the same in 2006. NetSuite did the same for the mid-market. The bundle was a constraint of the architecture, not a deliberate product choice.

Two things broke the constraint. First, accounting data moved out of the ledger and into upstream sources. Stripe pushed in revenue. Ramp and Brex pushed in card spend. Gusto and Rippling pushed in payroll. Mercury pushed in cash transactions. For a typical SaaS startup in 2025, much of the journal entry volume in QuickBooks comes from automated feeds. The ledger has become a destination. The system of record has moved upstream.

Second, language models got good enough to do the categorization and close work that used to require a person. Not perfectly, but well enough that a finance team of two can run a $100M ARR company.

The new bookkeeping layer

The first wave of AI-native entrants attacked the bookkeeping job specifically.

Kick was founded by Conrad Wadowski. After his $250M Teachable exit, he set out to build what Kick calls self-driving bookkeeping for the self-employed. The company raised a $9M seed from General Catalyst and the OpenAI Startup Fund in October 2024. The product reads bank and card transactions, identifies which ones are business-related, and categorizes them for tax. Cash-basis only, which limits the addressable market to sole proprietors and very small businesses. Inside that segment, Kick removes most of the work that QuickBooks Self-Employed still requires a human to do.

Puzzle.io took a different angle. Sasha Orloff, who previously built LendUp, is the CEO. He has raised more than $50M from General Catalyst and Felicis to rebuild the general ledger from scratch for startups on accrual accounting. The product pulls transactions automatically from the fintech stack (Stripe, Brex, and similar spend and payroll platforms) and lets the AI handle the categorization and entry work that used to require a junior bookkeeper. The pitch is real-time accrual financials for venture-backed companies that need clean books for fundraising.

Digits sits somewhere between the two. Jeff Seibert founded it in 2018 and launched it publicly in November 2023. The company is now sitting on close to $100M from Benchmark and other top-tier funds. The product calls itself an Autonomous General Ledger. It runs vertical LLMs trained on accounting data and ships AI agents that handle bill pay and categorization. Craig Walker joined the Digits leadership team in March 2025. He had previously co-founded Xero and served as its CTO. Reported customers include Particle News and Wispr.

These products differ in target customer and depth. They share an assumption: the bookkeeping work is mostly mechanical and can be done by software that reads transactions and applies rules. They are betting the data layer no longer needs a person in the loop for most companies.

The new accounting layer

The second wave is attacking the interpretation job. These products replace NetSuite.

Rillet was founded in 2024. The CEO is Nicolas Kopp, who had previously run N26 in the US. The company raised a $25M Series A from Sequoia in May 2025. A $70M Series B from Andreessen Horowitz and ICONIQ followed ten weeks later. The product is an AI-native ERP aimed at mid-market SaaS companies that have outgrown QuickBooks or Xero. These companies balk at a 12-month NetSuite implementation. Postscript closes its books in three days. Windsurf runs finance with two people. The close itself, the part that has always taken weeks of human judgment work, gets compressed by AI agents that reconcile entries and post accruals.

Campfire was founded by John Glasgow in 2023. He previously worked at Invoice2go, which Bill.com acquired for $625M. The company raised $100M across Series A and B inside twelve weeks in mid-2025, with Accel and Ribbit leading. The product replaces legacy ERPs with a general ledger and revenue automation, plus a conversational AI interface called Ember. Flex, one of Campfire's reference customers, cut its close from ten days to three with 67% fewer accounting resources.

DualEntry launched from stealth in October 2025. The $90M Series A was led by Lightspeed and Khosla. Co-founders Santiago Nestares and Benedict Dohmen had previously scaled an e-commerce company to nine-figure revenue on legacy ERP. They built DualEntry as an AI-native ERP covering general ledger, AR, AP, and FP&A in one system. The product leads with NextDay Migration, which moves historical data off legacy systems in 24 hours instead of the months that traditional ERP migrations take. Slash, a $140M revenue fintech, runs its entire finance function with one person on DualEntry. The migration was off QuickBooks Online. The platform automates roughly 90% of the manual tasks that team would otherwise be doing by hand.

These companies are going after the workload that Oracle and SAP have charged seven-figure ARRs to handle for decades. The close runs on AI agents with a small human team auditing the outputs.

What this means if you build software

If your product writes data into accounting software, the integration surface is widening. A customer on QuickBooks today might move to an AI-native ledger like Puzzle in two years. The data shape is similar across all of them because the underlying accounting model has not changed. The API mechanics, the auth flows, the field names, and the rate limits are all different. This is the integration problem that any accounting software integration team is now planning around.

Two things follow.

Integration surface area used to be an engineering question. It has become a product question. Picking one accounting integration to start used to be reasonable because QuickBooks dominated the US SMB market. Intuit's small business and self-employed segment generated $8 billion in revenue in 2023, and QuickBooks Online has more than 6 million customers globally. The dominance is real. The long tail is moving faster than it ever has, and the customers most likely to switch to an AI-native ledger are the same customers most likely to be early adopters of your product. If your integration strategy assumes QuickBooks plus Xero is enough, you will lose deals to competitors with broader coverage.

The AI-native ledgers are also designed API-first. They expose cleaner transaction data than QuickBooks does, with fewer of the historical quirks that come from a thirty-year-old data model. This is the practical face of open accounting. For products that read from accounting systems, this is good news. For products that write, it is more complicated, because each new ledger has its own validation rules and its own opinions about what a clean journal entry looks like.

The honest answer is that the accounting integration problem is getting harder. The number of platforms is growing and the rate of change inside each platform is increasing. Customers expect your product to work across all of them. Our accounting integration coverage radar shows what each platform supports and where the gaps are.

Apideck offers a unified accounting API. It connects to QuickBooks, Xero, and 20+ other ledgers through one integration, with consistent data models across vendors. The Apideck Accounting API is free for 30 days.