The announcement extends work Tipalti began in 2025 when it introduced a first generation of specialised AI agents. AP Agent represents a step change in scope: rather than automating discrete tasks, it is designed to orchestrate entire workflows, drawing on vendor history, general ledger data and company-specific accounting policies to make context-aware decisions rather than following static rule sets.
The product is built around what Tipalti describes as a configurable human-in-the-loop model. Finance teams can define which activities the agent executes autonomously and which require a human sign-off, with field-level reasoning surfaced to explain AI-generated values. Every action is logged in a full audit trail, and users can interrogate, clarify or override individual decisions through natural-language conversation. The company says the agent applies corrections from user feedback to future processing, improving accuracy over time.
Roby Baruch, chief product and technology officer at Tipalti, said the company’s aim is to give finance leaders “the confidence to reimagine what their teams and their businesses can achieve” as AI agents take on more autonomous responsibilities. Tipalti did not publish independent benchmark data for processing accuracy or straight-through reconciliation rates in the announcement.
The MCP addition is arguably the more structurally significant element. By implementing the Model Context Protocol, Tipalti allows its AP and payments data to be queried in real time through external AI tools without manual exports or platform switching. Paired with an ERP connection such as NetSuite , the capability can support month-end reconciliation, custom reporting dashboards and payment-term SLA tracking.
Source link







