Expensify launched 4 free AI integrations in 6 weeks as revenue sits flat at $33-$37 million a quarter

On October 15, Expensify▼−1.98% — Expensify, down 1.98 percent today (Nasdaq: EXFY) launched a native two-way integration↗ with Campfire, an AI-native ERP, the ledger system at the core of a company's books ¹. It is the expense vendor's third AI-native ERP plug-in in six weeks, after Rillet on September 1 and DualEntry on September 28, and on September 15 it also became a connector inside Anthropic'sPVT$350B — Anthropic, private, latest valuation $350B Claude for Small Business ²³⁴. One release is routine. Four moves in six weeks is a strategy: a flat-revenue expense vendor buying a route into the systems that AI-first finance teams are choosing.

Coding moves to the point of spend, and the month-end queue disappears

The integration is a two-way sync ¹⁵. Campfire↗'s chart of accounts flows into Expensify as categories and tags, so an expense gets coded where it happens rather than at month end. Out-of-pocket spend posts back as a vendor bill matched to the employee who submitted it, card spend posts as journal entries, and a daily sync carries reports to the ledger. Nick Tooker, Expensify's head of partnerships: Campfire customers "already run their close on AI", and now "nobody is exporting spreadsheets at month end" ¹. The capability to code a transaction was never the missing piece; the pipe was.

The connections are free because Expensify is buying distribution, not selling software

Expensify's revenue has sat between $33 million and $37 million every quarter for three years↗, with a loss in each ⁶. A vendor on that line does not price integrations; it pays for placement. The Claude connector, free on every paid Claude plan, is the same bet one layer up: be the tool an agent calls when an owner asks which receipts are missing, and the evening that question used to cost comes back ⁴. No source carries how many customers or dollars move through these pipes. The channel is real, and unmeasured.

Campfire's choice is sharper: own the agents, or rent the model

John Glasgow, Campfire's founder, CEO and CFO, put the dilemma on the record on October 13↗: "The CFO in me wants to pull back and lean on third-party AI like Claude. The CEO in me thinks... owning the AI agents and model is the only way to deliver on this vision" ⁷. His biggest 2027 bet is Ember, its AI product, funded by a $35 million Series A and then a $65 million Series B twelve weeks later ⁷. Owning the loop where the books live is the durable side of that trade, because the rented model is sold to every competitor on the same terms. The limit has moved anyway: Glasgow's rule, "If it can't be explained, a person reviews it before it posts" ⁷, now decides how far the close automates, not the model's ability to code.

Accountants learn the review loop; holders ask when the channel pays

The pipes are live; the shorter close is mostly still ahead, and Campfire's customer page points the way↗: ReplitPVT~$3B — Replit, private, latest valuation ~$3B scaled revenue without adding finance headcount ⁸. For accountants, the move is to learn the review loop, not to defend the keying. For EXFY holders the question is narrower: whether channel spend turns into growth before an agent makes the channel unnecessary.