JPMorgan Chase intends to roll out artificial intelligence agents later this year capable of operating autonomously for an hour or two at a stretch, a leap beyond current systems that typically run for only a few minutes, the bank told CNBC. The move offers collections and credit professionals an early look at how the largest U.S. bank by assets plans to put autonomous AI to work in regulated financial operations.
Derek Waldron, JPMorgan’s chief analytics officer, framed the shift as a transition from single-task tools to digital workers that manage entire workflows across multiple steps and separate software programs. He described the development as the start of an era of long-running autonomous agents, with the bank expecting to field them in 2026 once security and governance concerns are resolved. Over time, Waldron said, agents could remain coherent for hours, then days, then weeks.
For ARM operators weighing their own AI investments, the security framing is the notable part. Waldron acknowledged the technology is not yet ready for broad corporate use because of those concerns, a candid admission from an institution with a nearly $20 billion annual technology budget. That caution underscores the governance hurdles any collections firm will face before handing repetitive, multi-step processes to an agent that runs unsupervised.
JPMorgan also pointed to results it attributes to AI already in place. In private banking, systems that screen market activity, client positions and research overnight have contributed to a 20% increase in gross sales, according to the bank, which believes the tools could eventually let individual bankers expand client coverage by as much as half. Waldron argued that the strategic payoff is competitive advantage and revenue growth rather than maximum headcount reduction, though CEO Jamie Dimon has said some roles will be displaced and that affected staff will be retrained and redeployed.
A final point may resonate with technology vendors serving the collections space. Waldron said JPMorgan now examines more closely whether it can build capabilities internally rather than purchasing them, suggesting the competitive moat around certain software companies has weakened. For agencies that rely heavily on outside platforms, the calculus around building versus buying is one worth watching as agent capabilities mature.
