Software is now preparing corporate payments by the thousands without requiring a final click.
Coupa, a financial automation platform, announced its Payment Batch Creation Agent processed 2,395 individual payments totaling $20.1 million across 14 batches in its first five weeks. The tool handles every step except the final approval, which remains a human or separate system responsibility.
The line between preparation and authority
An AI agent reviewing an invoice or matching a purchase order differs from one deciding which approved invoices belong in a payment run. That change moves the software inside the transaction itself, not just the surrounding paperwork.
A Coupa customer said the agent cost about $27 in AI credits to operate and saved roughly $2,000 in staff time in one week. The platform also reported requisition cycle times fell by up to 50% for some users, though these figures come from self-reported data and may vary.
Most tasks now involve pattern recognition: validating invoices, confirming approvals, checking payment terms, and identifying early-payment discounts. Humans handle exceptions like suppliers updating bank details, unusually large invoices, or new payees.
This division of labor changes how payment networks manage trust when machines assemble transactions. If an agent makes a mistake, the issue isn’t just correction—it’s proving authorization existed in the first place.
How payment networks are adapting
Mastercard created a system called Verifiable Intent, which generates a tamper-resistant record of what a person authorized before the agent acts. Pablo Fourez, Mastercard’s chief digital officer, stated that as autonomy grows, trust must be demonstrated, not assumed. “If something goes wrong, everyone needs facts, not guesswork,” he said.
Visa is working on a similar Trusted Agent Protocol. The system lets merchants confirm they’re interacting with an approved agent and cryptographically links that agent to payment details. The objective remains consistent: ensuring actions can be traced back to a human decision, even if the person never directly handled the transaction.
Coupa addresses the challenge differently. Its latest release includes over 30 pre-built connections allowing third-party AI tools to access live financial data—procurement, invoicing, contracts, expenses—without custom integrations. Access follows the permissions of the employee who configured it, and every action stays auditable.
This approach points to a future where the agent preparing a payment might operate outside the software executing it. While $20.1 million represents a small portion of global corporate payment volumes, it’s no longer theoretical. A PYMNTS Intelligence report found 70% of firms already use at least one AI tool for cash flow management, with adoption expected to rise as more preparation tasks shift to software solutions.
The real test isn’t whether the agent can perform the work—it’s whether the system can verify accuracy and determine responsibility for errors.
