JPMorgan Chase plans to deploy artificial intelligence agents later this year that can operate autonomously for extended periods, a development CNBC has learned exclusively. The bank describes these agents as evolving from tools that complete single tasks to digital workers capable of managing workflows across multiple steps and various software programs.
Derek Waldron, JPMorgan’s chief analytics officer, said the new agents mark a shift toward long-running autonomous operations. “We’ve entered now the era of long-running autonomous agents,” he noted, explaining that agents can run for an hour or two rather than just a few minutes to accomplish a goal.
JPMorgan, led by CEO Jamie Dimon since 2006, is the largest U.S. bank by assets and operates with a technology budget approaching $20 billion annually. While much of AI discussion centers on model intelligence, Waldron said the key question for enterprises is how long AI systems can operate effectively without human intervention. He described this as “intellectual coherence,” enabled by improvements in AI reasoning that position agents as more of a team manager than a single worker—capable of delegating activities and sustaining more complex operations.
New capabilities advancing agents include writing code, controlling web browsers, and interacting with desktop software. Although long-running agents are not yet ready for broad corporate deployment due to security concerns, Waldron projected that such capabilities will arrive in 2026.
Observationally, AI agents are expected to stay coherent across extended periods—multiple hours, then days, and eventually weeks, Waldron said. The trajectory suggests these tools could increasingly impact revenue-generating roles, not only back-office or development tasks.
In private banking, for example, AI systems can screen market activity, monitor client positions, and review research overnight, enabling bankers to concentrate more on client interactions. Waldron cited a 20% rise in gross sales attributed to these tools and asserted that individual bankers might expand their client reach by as much as 50% as deployment scales.
Dimon has acknowledged that AI could displace some workers, and JPMorgan has said it will train and redeploy employees affected by the changes. Yet Waldron stressed that firms are shifting from viewing AI as a cost-cutting instrument to recognizing its potential for expanding revenue. “For enterprises to win with AI, it’s not about cutting the maximum number of jobs,” he said, adding that sustainable competitive advantage is the broader aim.
The bank’s strategy around building versus buying software has also evolved, with JPMorgan now weighing in-house development as a potential path that could pressure traditional external vendors. Waldron’s remarks underscore the bank’s broader view that AI-driven productivity gains can extend beyond automation of simple tasks to supporting more complex, revenue-related activities.
