Paper still defines payments’ last mile. J.P. Morgan Payments thinks AI and robotics can tackle that.

The dominant narrative in payments is real-time payments, stablecoins, instant payment rails, always-on cash flow, and near-seamless cross-border movement. But speed at the point of settlement doesn’t solve what happens before and after money moves.

One of the world’s largest payments businesses processed more than 130 million checks in 2025. Behind every check was a chain of envelopes, remittance slips, invoices, inconsistent formats, and manual reconciliation.

That’s the gap J.P. Morgan Payments is targeting with its multi-year effort, using AI, robotics, computer vision, and large language models to digitize and automate its lockbox operations. The project is about modernizing one of the most manual information-processing systems still embedded inside corporate payments.

“Checks are still a meaningful part of the U.S. payments landscape,” says Michelle Conklin, Head of Receivables and Public Sector at J.P. Morgan Payments. “The challenge is often not the check payment itself, but all of the manual work that comes with it: opening mail, extracting checks and remittance documents, capturing data, validating information, and handling exceptions.”

“That is where we see a real opportunity to drive value,” she notes. “When we automate those steps with AI and robotics, we can scale more easily and make the process faster and more accurate. This is a reminder that innovation in payments is not only about faster settlement. It is also about removing friction from the end-to-end process, such as what we’re advancing in the lockbox space.”

When AI finally became good enough for the messy middle

…