JPMorgan plans to deploy long-running AI agents that can work for hours without human input

For the past two years, the AI race has focused on smarter models. The next phase may be about something far more consequential: how long AI can work on its own before it needs a human to step in.

JPMorgan Chase is preparing to deploy a new generation of artificial intelligence agents later this year that can operate autonomously for far longer than current systems, marking a significant step in enterprise AI adoption.

The bank’s chief analytics officer, Derek Waldron, told CNBC that AI agents are moving beyond simple task completion and becoming digital workers capable of managing entire workflows across multiple applications and systems.

“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.”

The comments offer a glimpse into where large enterprises see AI heading next. Much of the public conversation has centered on model performance, reasoning capabilities, and benchmark scores. Inside major corporations, attention is shifting to a different metric: how long AI systems can remain productive and coherent without human oversight.

JPMorgan’s interest carries weight. Led by CEO Jamie Dimon since 2006, the bank is the largest U.S. bank by assets and spends nearly $20 billion annually on technology. A deployment at that scale signals growing confidence that autonomous AI systems are approaching a level suitable for use inside highly regulated organizations.

From AI Assistants to Autonomous Digital Workers

The idea of long-running agents has gained momentum over the past year. Tools such as Anthropic’s Claude Code and OpenClaw have demonstrated that AI can handle extended tasks involving coding, research, and software interaction. Large enterprises have been slower to adopt such systems, largely due to security, governance, and reliability concerns.

Waldron believes the technology is approaching an inflection point.

Much of that progress comes from improvements in reasoning capabilities. He described the concept as “intellectual coherence,” referring to an AI system’s ability to stay focused on a goal over extended periods without losing context or making costly mistakes.

“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.

The latest generation of AI agents can already write code, control web browsers, and interact directly with desktop software. Those capabilities allow them to perform work that previously required constant human guidance.

Security remains one of the biggest barriers before enterprises fully embrace autonomous agents. Waldron acknowledged that long-running systems are not yet ready for broad corporate deployment but suggested that the timeline is shortening.

“We will have those in 2026.”

He expects the technology to continue advancing beyond hour-long sessions.

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

Why JPMorgan Is Betting on Long-Running AI Agents

The productivity gains from AI are already becoming visible across JPMorgan. Software development and back-office operations have been among the earliest beneficiaries, though Waldron said the impact is increasingly showing up in revenue-generating roles.

Private banking offers one example. AI systems now analyze market activity, client positions, and research overnight, giving bankers a clearer picture before speaking with clients the next day.

According to Waldron, those tools have contributed to a 20% increase in gross sales. The bank believes AI could eventually enable individual bankers to expand client coverage by up to 50%.

The rise of AI agents is expected to reshape parts of the workforce. Dimon has previously acknowledged that some jobs at JPMorgan will be displaced by AI, adding that the company intends to retrain and redeploy affected employees.

Waldron said many organizations initially viewed AI through the lens of cost reduction. That mindset is beginning to shift as companies discover new ways to grow revenue and improve customer outcomes.

“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.”

The shift is also influencing software purchasing decisions within the bank.

JPMorgan is taking a closer look at which capabilities it can build internally rather than buy from third-party vendors. That change could create new pressure for software companies whose products rely on features that AI can increasingly replicate.

“The moat around certain types of software companies is most certainly diminished versus where it was in the past,” Waldron said.

If JPMorgan’s timeline proves accurate, the next chapter of enterprise AI may not be defined by smarter chatbots. It may be defined by autonomous digital workers that operate for hours, and eventually days, with minimal human involvement.