JPMorgan plans longer-running AI agents for corporate use this year
The bank says new AI agents could work for hours across tasks, while existing tools are already lifting private banking sales.
By Daniel Okafor · Business Editor
3 min read
JPMorgan Chase plans to introduce more advanced artificial intelligence agents later this year that can work on tasks for much longer stretches than current systems, CNBC reported. The move signals that one of the largest corporate AI users sees the technology getting closer to broader use inside heavily regulated companies.
Derek Waldron, JPMorgan’s chief analytics officer, told CNBC that AI agents are moving beyond single-task tools into systems that can coordinate work across several steps and multiple software programs. He said the next generation of agents can operate for an hour or two, compared with systems that typically run for only a few minutes before needing more human direction.
Waldron described the shift as the start of “long-running autonomous agents,” according to CNBC. He said the bank expects to have those capabilities in 2026, though he also said such systems are not yet ready for broad corporate use because security issues still need to be addressed.
Agents move from tasks to workflows
According to Waldron, the key test for the technology is how long an AI system can keep working effectively before a person has to step in. He called that ability “intellectual coherence” and said recent improvements in AI reasoning have helped agents act more like managers that break a problem into smaller jobs.
Waldron told CNBC that newer agents can also write code, operate web browsers and interact with desktop software, which allows them to handle more complex assignments. He said he expects the systems to remain coherent for longer periods over time, moving from multiple hours to days and eventually weeks.
CNBC reported that long-running agents have gained attention over the past year through products and projects including Anthropic’s Claude Code and OpenClaw. JPMorgan’s plans suggest the tools are nearing the point where large companies may be able to meet internal security and governance requirements for deployment.
JPMorgan, led by Chief Executive Jamie Dimon since 2006, is the biggest U.S. bank by assets, according to CNBC. The company has a yearly technology budget of nearly $20 billion, giving it the scale to build and test AI systems across a large workforce.
Bank says AI is lifting sales
Waldron told CNBC that AI has already produced business gains at JPMorgan beyond software development and back-office work. In private banking, he said, AI tools review market activity, client holdings and research overnight so bankers can spend more time with clients.
The bank has recorded a 20% increase in private banking gross sales tied to those tools, Waldron said. He also told CNBC that JPMorgan believes AI could eventually let individual bankers expand client coverage by as much as 50%.
Dimon has said some employees will be displaced by AI and that JPMorgan is preparing to train and redeploy affected workers, CNBC reported. Waldron said companies are also beginning to view AI as a way to increase revenue rather than only as a way to reduce costs.
Waldron said JPMorgan’s view of whether to build software internally or buy it from vendors has changed as AI capabilities improve. He told CNBC that the bank is examining more closely where it can develop tools itself, and said the defensive position of some software companies has weakened compared with the past.
This story draws on original reporting from CNBC.