AI Agents Now Handle Corporate Bill Pay
Visa and Ramp Want Your AI Agent to Pay the Bills
Ramp and Visa are putting AI agents in charge of corporate bill pay.
The two companies announced an expanded partnership Tuesday that introduces AI agents designed to automate corporate bill payments, reduce manual finance-team workflows, and enforce real-time spend controls across Ramp's customer base of more than 50,000 businesses. The agents will plug into Visa's global payments network through the Visa Intelligent Commerce platform and the Visa Trusted Agent Protocol — Visa's framework for authorizing and securing machine-to-machine payment transactions.
The announcement is light on technical detail, but the architecture is clear enough: instead of a human initiating a wire or ACH transfer, a software agent handles the workflow end-to-end, from invoice matching to payment execution, with built-in controls that flag or block transactions outside policy. Visa's protocol handles authentication and authorization at the network layer. Ramp handles the approval workflows, expense policies, and bookkeeping.
"The best financial systems don't add controls after the fact — they build them into every transaction," said Colin Kennedy, chief business officer at Ramp, in a press release on PR Newswire. "That's what we're delivering with Visa."
Chris Newkirk, president of commercial and money movement solutions at Visa, said in the same release: "Enterprises are looking for payment solutions that reduce friction, not add to it. Ramp's approach to automation and real-time controls aligns with Visa's mission to make commerce simpler and more secure."
The partnership builds on an existing multi-year issuing relationship. Ramp's customers use Visa co-branded corporate cards, but this expansion goes further: Visa will also use Ramp for corporate service use cases — a sign that the integration is bidirectional and that Visa sees Ramp's platform as infrastructure worth building on, not just a card program to issue.
Ramp's enterprise customer base grew 133% year over year in 2025, per Kennedy's blog post. The company says it processes over $100 billion in purchases annually and has saved customers a combined $10 billion and 27.5 million hours since its 2019 founding.
The agentic angle is real, but the "autonomous finance" framing in the press release title deserves scrutiny. Corporate bill pay involves vendor relationships, approval hierarchies, and audit trails that exist for legal and regulatory reasons — not because humans are slow. Automating the execution is tractable. Automating the judgment calls is where agentic AI in finance runs into its actual ceiling.
That said, the use case for fully automated low-value, high-volume payments — subscriptions, recurring vendor invoices, fleet card transactions — is legitimate and not glamorous. It's exactly the kind of workflow where adding an AI agent doesn't require explaining why it exists. The ROI is obvious when the alternative is a finance associate processing 200 invoices a week.
This is distinct from consumer AI paying your bills on your behalf. Corporate spend has compliance requirements, policy controls, and multi-stakeholder approval chains that consumer finance doesn't. The agent works within that structure, not around it. Whether the Visa Trusted Agent Protocol enforces those controls robustly enough for a CFO to sign off is the material question — and one the press release doesn't answer.
Ramp and Visa aren't alone in this lane. Visa said last year that it had already processed "hundreds of controlled, real-world agent-initiated transactions" in partnership with Skyfire, Nekuda, PayOS, and others. The Ramp announcement adds scale — 50,000 businesses on a platform that already handles their corporate card spend — to a pattern that was already live in Visa's ecosystem.
For founders and builders: the infrastructure for agent-initiated B2B payments exists. The question is which vertical SaaS platforms will be first to wire it into their workflows. Ramp is moving fast on this with Visa, which suggests the answer for enterprise spend management is sooner rather than later.
Primary sources:
† Add footnote: "Source-reported; not independently verified."
†† Add footnote: "Source-reported; not independently verified."