JPMorgan’s consumer chief says agentic commerce will take time, citing trust, security and payment preferences

JPMorgan Chase is moving cautiously toward “agentic” commerce, the idea that artificial-intelligence agents do more than summarize products or help customers find options, and instead take action on a consumer’s behalf. In comments at a Morgan Stanley investor conference, Diane Lake, the bank’s CEO of consumer and community banking, said the industry has already adopted AI tools for search and discovery because they reduce friction. But she said the leap from planning and shopping to completing transactions is a different matter.

Lake said JPMorgan is not seeing a broad shift of AI systems into the transaction step of online purchasing, where real funds move and errors carry greater consequences. She suggested that this lag is likely to persist for some time. “Because when people are moving money, things change,” she said, adding that trust and security become more important in that phase of the customer journey.

The bank’s caution is also tied to customer protections and consumer payment preferences. Lake said JPMorgan is working to keep core consumer payment choices, including common methods such as cards and digital wallets, connected to agent-enabled commerce. She emphasized that protections should remain in place even when an AI agent is performing actions behind the scenes.

A key condition, Lake said, is keeping humans in the loop for “things that matter,” along with transparency about what the agent is doing. She also pointed to the need for an operational and legal “liability framework” that would work if an agent makes a mistake. In her view, addressing those governance and responsibility issues is essential before delegating more purchase steps to automated systems.

While she was skeptical about consumers delegating general purchasing decisions to agents “just yet,” Lake drew a sharper contrast with travel planning and booking. She said travel is still complex, even with AI help, because it requires coordinating multiple components such as flights, hotels, itineraries, and changes along the way. That complexity, she noted, makes it harder to manage disputes and alterations when something does not go as planned.

JPMorgan’s near-term focus, according to Lake, is not broad consumer purchasing by autonomous agents, but a more constrained test case. She said the company is aiming for a consumer-facing AI travel agent pilot before the end of the year. In a travel context, the bank appears to be looking for areas where an agent can reduce planning work while still controlling risks through guardrails and supervised decision-making.

The conversation highlights a wider challenge for banks and consumer tech alike as AI capabilities expand. It is relatively straightforward to use AI for product discovery, comparisons, and summaries. Moving into the transactional layer adds interlocking concerns, including authentication, fraud and dispute handling, payment authorization rules, and who is responsible when automated actions go wrong.

JPMorgan did not provide additional details in the cited conference remarks about specific timelines beyond the planned end-of-year pilot, the consumer channels involved, the scope of what the agent would be allowed to book, or whether the pilot would include full payment authorization or human confirmation at each step. The bank also did not disclose any quantitative metrics on expected adoption or performance, or how it intends to measure safety and consumer satisfaction for agent-driven travel planning.

Investors and customers will likely watch for how JPMorgan structures the pilot, particularly around the “humans in the loop” operating model and the liability approach Lake referenced. Future updates could clarify what kinds of travel changes the agent can handle, how disputes or refunds are managed, and whether JPMorgan expands beyond travel after demonstrating that protections can be preserved in agentic workflows.