J.P. Morgan eliminates 13 billion keystrokes with AI lockbox

Automating the paper trail behind B2B payments

J.P. Morgan Payments processes roughly 480 million checks and payment documents annually through its lockbox network. Each payment arrives with invoices, remittance slips, handwritten notes, and formatting variations that historically demanded manual data entry. Before automation, employees performed approximately 13 billion keystrokes every year just to process these documents.

Rather than wage war on checks—which still represent 25% to 26% of U.S. B2B payments—the bank focused on eliminating the labor they create.

In 2020, J.P. Morgan rebuilt its lockbox platform with AI embedded in core workflows. Computer vision and machine learning now extract payment information, validate business rules, and review documents automatically after scanning. The bank recently added large language models to handle increasingly complex exceptions.

The system now processes more than 4,000 envelope and document variations while achieving over 99.999% accuracy in data extraction and business rule validation, according to details first reported by Tearsheet.

Physical automation meets digital intelligence

In 2025, J.P. Morgan extended automation into the physical realm by deploying robotics at its lockbox facility. These systems open envelopes, extract checks and invoices, unfold documents, organize paperwork, and prepare everything for AI processing.

"By investing in robotic and AI technology to improve our lockbox operations, we are automating the most labor-intensive tasks of the process, freeing our team to focus on more complex, higher-value decision-making," said Michelle Conklin, Head of Receivables and Public Sector at J.P. Morgan Payments.

The robots launched at a single site for testing. Following successful deployment, they're being refined and will return later this summer as part of a phased rollout.

Why it matters

Settlement speed gets attention, but operational friction happens in the workflows surrounding payments. For treasury and finance teams, payments must be matched to invoices, reconciled against receivables, and reflected in accounting systems before they become useful. AI that can interpret thousands of document variations and extract meaning from unstructured data changes the economics of processing legacy payment formats. This matters because businesses built decades of workflows around paper—eliminating those workflows entirely isn't realistic, but making them machine-readable is now economically viable.

Beyond moving money

"Moving dollars is only half the story," Conklin noted. The other half ensures payment data is accurate and actionable when funds arrive, helping businesses reduce days sales outstanding, improve working capital, and accelerate reconciliation.

Checks persist because they're embedded in established business processes. What's shifting are the costs and capabilities of processing them. AI has reached a threshold where it can automate work that previously required human judgment at scale.

The biggest productivity gains in financial services may come not from eliminating legacy payment methods, but from making them machine-readable. These details were first reported by Tearsheet.