The bank will deploy more powerful agents this year

JPMorgan Chase plans to roll out artificial intelligence agents later this year that can operate autonomously for far longer than existing versions, marking a new milestone in enterprise adoption of AI, CNBC has exclusively learned.

AI agents are evolving from tools that complete single tasks to digital workers that manage multi-step workflows and disparate software, Derek Waldron, JPMorgan’s chief analytics officer, told CNBC in an interview.

“We have now entered the era of long-term autonomous agents,” Waldron said. This “means that agents don’t just run for two or three minutes to carry out an objective or some instructions from a human, they can run for an hour or two.”

Longtime agents have already emerged over the past year as examples, including Claude Code and Anthropic’s OpenClaw, have gone viral. JPMorgan’s planned rollout, however, suggests the technology is close to overcoming the security and governance hurdles that have slowed its adoption among large enterprises.

JPMorgan, led by CEO Jamie Dimon since 2006, is the largest U.S. bank by assets and has an annual technology budget of nearly $20 billion.

While much of the debate around generative AI has focused on the intelligence of models, technology leaders are increasingly focusing on a different question, Waldron said: How long can AI systems operate effectively before requiring human intervention?

This concept, which Waldron called “intellectual coherence,” was helped by improvements in the way AI models reason, allowing them to be more of a “team leader than an individual worker,” he said.

“Just like how people operate, team leaders can analyze a problem and delegate activities, and teams can operate much longer to do more complex things,” Waldron said.

Other recent advances that have helped agents perform more complex tasks include the ability to write code, control web browsers and interact directly with desktop software, he said.

Although the longtime agents aren’t yet ready for business use due to security concerns, their arrival isn’t far away, Waldron said: “We’ll have them in 2026.”

Eventually, AI agents will remain consistent for “several hours, then days, then weeks,” he said.

Productivity gains from AI have been most visible in software development and back-office type operations, but Waldron said they are increasingly bolstering revenue-generating roles.

In private banking, for example, AI systems monitor market activity, client positions and searches overnight, helping bankers focus on client interactions.

The bank has seen a 20% increase in gross sales from these tools, he said, and estimates they could ultimately allow individual bankers to expand their customer coverage by up to 50%.

Dimon made it clear that some of his employees will be displaced by AI, saying the company is preparing to train and redeploy employees affected by the changes.

But Waldron added that while many companies initially looked at AI as a cost-cutting tool, they are increasingly recognizing its potential to increase revenue.

“For businesses to win with AI, it’s not about cutting as many jobs as possible,” he said. “It’s all about trying to create a sustainable competitive advantage.”

Waldron said the bank’s thinking about building rather than buying software from outside vendors has also changed. JPMorgan is now taking a closer look at its ability to develop capabilities in-house, he said, which could put pressure on some traditional suppliers.

“The gap around certain types of software companies is most certainly narrowed compared to what it was in the past,” he said.

— CNBC’s Gabrielle Fonrouge contributed to this report.