JPMorgan Chase plans to deploy more powerful AI agents later this year that can operate autonomously for longer periods, raising a major enterprise question about whether large companies are ready to trust agents with real business workflows.

The bank’s plan moves the AI agent conversation beyond chatbots and copilots. Derek Waldron, JPMorgan’s chief analytics officer, told CNBC that agents are evolving from tools that complete single tasks into digital workers that manage workflows across multiple steps and software programs.

Enterprise leaders, risk teams, technology vendors, and workers now face a practical deployment decision over which workflows can be trusted to autonomous agents, how long those agents should operate before review, and what oversight large companies need before agents act inside real business 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.”

In short, JPMorgan is preparing for AI agents that can work for hours, turning enterprise AI from a productivity tool into a question of workflow execution, oversight, and trust. The bank’s plans suggest large companies are beginning to evaluate where autonomous agents can perform meaningful work, and where human control still needs to remain close.

Long-running AI agents are autonomous AI systems designed to carry out multi-step workflows over extended periods before requiring human intervention.

JPMorgan’s long-running AI agents are planned autonomous systems designed to work across multi-step enterprise workflows for longer periods before human intervention.

JPMorgan Chase plans to deploy more powerful AI agents later this year that can work autonomously for far longer than existing versions, CNBC reported.

Derek Waldron, JPMorgan’s chief analytics officer, described the development as the beginning of the era of long-running autonomous agents. 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,” he said.

The longer operating window changes what enterprises can ask an agent to do. Waldron said AI agents are evolving from tools that complete single tasks into digital workers that manage workflows across multiple steps and separate software programs.

That moves the focus from conversational assistance toward workflow execution. A long-running agent can take a goal from a human and work across several steps before requiring intervention, which makes trust, access, oversight, and operational risk part of the deployment decision.

As the largest U.S. bank by assets, with a nearly $20 billion annual technology budget and CEO Jamie Dimon at the helm since 2006, JPMorgan would be deploying long-running agents inside a highly complex enterprise environment in finance. In that setting, task completion would be only one part of the deployment test. Long-running agents would also need clear controls for system access, security, human oversight, and operational risk.

Waldron said technology leaders are increasingly focused on how long AI systems can operate effectively before they need human intervention.

He called that concept “intellectual coherence.” In practical terms, the question is whether an AI system can keep track of a goal, break the work into parts, use the right tools, and continue making useful progress over time.

The key point: long-running agents are different because they are designed to stay coherent across extended workflows, multiple tools, and several steps of execution.

Waldron compared the change to the way a team manager works. “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,” he said.

That comparison explains why long-running agents could be more useful inside large companies. Many enterprise workflows are not single-step tasks. They involve checking information, moving between systems, coordinating actions, reviewing results, and escalating issues when something falls outside the expected path.

Waldron said recent advances have helped agents do more complex jobs, including the ability to write code, control web browsers, and interact directly with desktop software. Those capabilities give agents more ways to act inside existing business environments.

Waldron said long-running agents are not yet ready for corporate use because of security concerns.

For enterprises, that caveat turns long-running agents into a trust and control question. Agents that operate for an hour or two can move beyond single-task assistance, which means companies need stronger controls around access, monitoring, permissions, human review, error handling, unauthorized action, data exposure, and flawed execution.

Waldron said the arrival of long-running agents is not far off. “We will have those in 2026,” he said.

He also described a longer-term path in which agents remain coherent for “multiple hours, then days, then weeks.” For large companies, that progression depends on whether security and governance can keep pace with longer autonomy. JPMorgan expects long-running agents in 2026, but larger corporate deployment will depend on whether security and governance can support longer autonomy.

Waldron said AI-driven productivity gains have been most visible in software development and back-office operations, but AI is increasingly helping revenue-generating roles.

In private banking, Waldron said AI systems screen market activity, client positions, and research overnight. That work helps bankers focus more time on client interactions, and JPMorgan has seen a 20% increase in gross sales because of these AI tools, he said. He also said the tools could eventually allow individual bankers to expand client coverage by as much as 50%.

As AI takes on more work inside JPMorgan, the impact extends to workforce planning. Dimon has said some workers will be displaced by AI, and the firm is preparing to train and redeploy employees affected by the changes.

Waldron said many companies initially approached AI as a cost-cutting tool, but 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 JPMorgan’s thinking around building versus buying software from outside vendors has also changed. The bank is looking more closely at whether it can build capabilities in-house, he said, potentially 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,” Waldron said.

Across revenue, workforce, and software decisions, JPMorgan is treating AI as an operating strategy, not just a productivity tool. The larger enterprise test is whether companies can trust AI agents to execute longer workflows safely across business systems.

Q: What is JPMorgan doing with AI agents?
A: JPMorgan Chase plans to deploy more powerful AI agents later this year that can work autonomously for longer than existing versions. Derek Waldron told CNBC that these agents can run for “an hour or two,” rather than only two or three minutes.

Q: What makes JPMorgan’s AI agents different from copilots?
A: Long-running AI agents are designed to manage multi-step workflows across software programs and continue working for longer periods before human intervention. A copilot typically assists a human with a task. A long-running agent can take a goal, break it into steps, use tools, and continue executing the workflow over time.

Q: Why should large companies care about JPMorgan’s AI agent plans?
A: JPMorgan’s plan raises the issue of enterprise trust in autonomous agents. Large companies need to know whether agents can stay coherent, secure, and accountable while operating inside real business workflows.

Q: Why can’t JPMorgan use long-running AI agents broadly yet?
A: Security concerns still limit corporate deployment of long-running agents. Waldron said the agents are not yet ready for corporate use, even though he expects JPMorgan to have them in 2026.

Q: How is JPMorgan already using AI in private banking?
A: JPMorgan says AI tools in private banking screen market activity, client positions, and research overnight, helping bankers focus on client interactions. Waldron said the bank has seen a 20% increase in gross sales because of these tools and believes individual bankers could eventually expand client coverage by as much as 50%.

Q: What could JPMorgan’s AI strategy mean for jobs and software vendors?
A: JPMorgan’s AI plans could affect workforce redeployment and enterprise software buying decisions. Jamie Dimon has said some workers will be displaced by AI and retrained or redeployed, while Waldron said JPMorgan is looking more closely at building some capabilities in-house.

JPMorgan’s plans turn AI agent deployment into an operating decision, not only a software adoption decision. If agents can run for an hour or two, companies need clear rules for what those systems can access, what actions they can take, and when humans must review the work.

The main enterprise issue is that agent duration changes the risk profile. A single-task AI assistant can produce an output for a person to review. A long-running AI agent may move through multiple systems and steps before review, making coherence, security, and accountability central to enterprise use.

Enterprise leaders, technology teams, risk officers, compliance teams, software vendors, and workers should pay attention because JPMorgan is testing a question many large companies are likely to face as agents gain longer operating windows. If a company with JPMorgan’s scale and technology budget is preparing agents that can run for hours, other enterprises will need to evaluate whether their own workflows, controls, and vendor strategies are ready for more autonomous AI systems.

The near-term issue is preparation. Waldron expects JPMorgan to have long-running agents in 2026, while security concerns still limit corporate use. That gives large companies a planning window to decide which workflows are suitable for autonomous agents, what controls need to be in place, and where human review remains necessary.

As AI moves from employee assistance and single-task tools toward autonomous workflow execution, companies need to decide which workflows can be trusted to agents, how long those agents should operate before review, and where human judgment must remain in control.

In short, JPMorgan’s long-running AI agent plans point to a more advanced stage of enterprise AI adoption. The harder question is whether companies can trust autonomous agents with workflows that affect customers, workers, revenue, and operational risk.

The future of enterprise AI agents will be decided less by how impressive they look in demos and more by whether companies can trust them to keep working safely when no one is watching every step.

Editor’s Note: This article was created by Alicia Shapiro, CMO of AiNews.com, with writing support, AEO/GEO/SEO optimization, image concept development, and editorial structuring support from ChatGPT, an AI assistant. All final editorial decisions, perspectives, and publishing choices were made by Alicia Shapiro.