Ramp Launches AI Agents to Automate Corporate Procurement

AI agents can research products and negotiate purchases, but allowing software to spend corporate money creates risks around overspending, credential theft and accountability. Ramp is positioning its financial operations platform as a control layer for this emerging agent economy, giving tools such as ChatGPT and Claude payment authority while preserving company budgets, merchant restrictions, approval chains and an auditable record of each action.

Ramp launched Agent Cards on March 11, 2026, issuing a fresh PAN and CVV for a single merchant and capping it at the approved purchase amount. The credential expires after its first authorization or 12 hours. Available through API, Model Context Protocol and the Ramp CLI, the service uses four controls — identity, policy, capabilities and audit trail — to carry an agent-led purchase from checkout through receipts, memos and accounting classification.

Corporate travel and expense management has traditionally required employees to assemble receipts, code transactions and wait for managers and finance teams to review claims. SAP Concur is using Joule to coordinate specialized AI agents that can initiate reports, extract missing receipt data and check submissions against company policy. The shift moves spending controls closer to the point of purchase, reducing rework while giving finance teams earlier visibility into costs and compliance risks.

SAP Concur on Aug. 6, 2026 unveiled additional tools spanning bookings, approvals, meeting planning, expense oversight and disruption handling, including capabilities developed with Amex GBT. An AI-assisted delegate dashboard, initially available to early adopters, is designed to cut issue-resolution time by as much as 25% by consolidating traveler profiles, approvals and expense reports. Other agents analyze receipts and flag policy violations before submission. SAP Concur did not disclose pricing or an investment amount.

Procurement teams have traditionally relied on manual data entry and fragmented handoffs to move a purchase from request through invoice payment, creating delays and reconciliation risks. Procurify, which targets mid-market organizations, is positioning company-specific spending, purchasing and vendor data as the context that lets AI agents act inside existing controls. The shift matters as roughly 73% of enterprises are using or considering generative AI to improve procurement, moving competition beyond analytics toward software that can execute routine workflows.

Procurify said on July 20, 2026, that it completed the conversion of its platform into a fully agentic procurement system by adding three purpose-built agents. Guided Intake uses a conversational interface to walk employees through policy-checked purchase requests. Order Autopilot draws on an organization’s purchasing history to code line items as orders are created, while a rebuilt agent-powered accounts payable engine matches and codes invoices against purchase and vendor records. The company said the agents learn from each customer’s business context and improve with use.