- JPMorgan Chase plans to deploy AI agents later this year that can operate autonomously for hours at a time,
- AI agents are evolving from tools that complete single tasks to digital workers that manage workflows across multiple steps and disparate software programs, Derek Waldron, JPMorgan chief analytics officer,
- JPMorgan says AI is already boosting business results, including a 20% increase in private banking gross sales, and could eventually allow bankers to expand client coverage by as much as 50%.
JPMorgan Chase plans to deploy artificial intelligence agents later this year that can work autonomously for far longer than existing versions, marking another milestone in the corporate adoption of AI,
AI agents are evolving from tools that complete single tasks to digital workers that manage workflows across multiple steps and disparate software programs, Derek Waldron, JPMorgan chief analytics officer,
“We’ve entered now the era of long-running autonomous agents,” Waldron said. That “means that agents don’t just run for two or three minutes to carry out a goal or some instructions of a human, they can run for an hour or two.”
Long-running agents have already emerged over the past year as examples including Anthropic’s Claude Code and OpenClaw went viral. JPMorgan’s planned deployment, however, suggests the technology is close to clearing the security and governance hurdles that have slowed adoption inside large companies.
JPMorgan, run by CEO Jamie Dimon since 2006, is the biggest U.S. bank by assets and has a nearly $20 billion annual technology budget.
While much of the conversation around generative AI has focused on model intelligence, tech leaders are increasingly focused on a different question, said Waldron: How long can AI systems operate effectively before requiring human intervention?
That concept, which Waldron called “intellectual coherence,” has been helped by improvements in how AI models reason, enabling them to be more of a “team manager than an individual worker,” he said.
“Just like how people function, team managers can parse out a problem and delegate activities, and teams can run for a lot longer to do more complex things,” Waldron said.
Other recent advances that have helped agents do more complex jobs include the ability to write code, control web browsers and interact directly with desktop software, he said.
While long-running agents aren’t yet ready for corporate use because of security concerns, their arrival isn’t far off, Waldron said: “We will have those in 2026.”
Eventually, AI agents will remain coherent for “multiple hours, then days, then weeks,” he said.
‘Diminished’ moats
AI-driven productivity gains have been most visible in software development and back-office type operations, but Waldron said it is increasingly boosting revenue-generating roles.
In private banking, for example, AI systems screen market activity, client positions and research overnight, helping bankers focus on client interactions.
The bank has seen a 20% increase in gross sales because of these tools, he said, and believes they could eventually allow individual bankers to expand client coverage by as much as 50%.
Dimon has been clear that some of his workers will be displaced by AI, saying that the firm is preparing to train and redeploy employees impacted by the changes.
But Waldron added that while many companies initially approached AI as a cost-cutting tool, they are increasingly recognizing its potential to expand revenue.
“For enterprises to win with AI, it’s not about cutting the maximum number of jobs,” he said. “It’s all about trying to create a sustainable competitive advantage.”
Waldron said that the bank’s thinking around building versus buying software from outside vendors has also shifted. JPMorgan now looks more closely at whether it can build capabilities in-house, he said, possibly putting pressure on some traditional vendors.
“The moat around certain types of software companies is most certainly diminished versus where it was in the past,” he said.
