Tallyfi raises $58 million Series B for AI agents that police corporate spending in real time
The round values the corporate-card and expense platform at roughly $420 million and will fund its point-of-transaction policy engine.
Tallyfi, a corporate card and spend-management platform, has raised $58 million in a Series B round led by Ribbit Capital, with Thrive Capital, Foundation Capital and existing investor Homebrew participating. The round implies a valuation of roughly $420 million and follows a $19 million Series A closed last year.
The company’s pitch is about where enforcement happens rather than what it enforces. Conventional expense software flags a policy violation after the fact, in a monthly report an employee has to reconcile. Tallyfi’s AI agents sit at the point of transaction instead, approving, declining or routing a purchase for manual review in real time based on a company’s spend rules, remaining budget and vendor risk signals.
The product issues virtual and physical corporate cards, auto-categorizes and codes transactions to the general ledger as they post, and generates expense reports without requiring an employee to submit one — collapsing a workflow that has traditionally run through several separate tools and a manual close-of-month reconciliation.
The category is not empty. Ramp, Brex, Airbase and Navan all sell some version of automated spend control, and each has added AI-driven categorization and anomaly detection over the past two years. Tallyfi’s differentiation rests on timing — intercepting a purchase before it happens rather than auditing it afterward — which is a real product distinction but one competitors with larger balance sheets can attempt to replicate.
The business depends substantially on card interchange revenue, a model that is sensitive both to interchange-fee regulation and to how much of a customer’s total spend actually runs through the card rather than ACH or wire. As larger incumbents cut card economics to defend share, newer entrants face margin pressure on the same revenue line their growth depends on.
The round extends a pattern that has shown up elsewhere in fintech funding this year, including at accounting-software company Rillet: AI agents moving earlier into the workflow, embedded at the point a transaction occurs rather than layered on top of records after the fact. It is the same underlying thesis — that finance software built before large language models exist to review work is structurally behind software built to prevent errors before they’re recorded — applied to a different back-office function.