JPMorgan Plans to Deploy More Powerful AI Agents This Year, Capable of Running Autonomously for One to Two Hours
Alina Collins
JPMorgan plans to deploy a new generation of AI agents later this year that can work autonomously for one to two hours — evolving from single-task tools into cross-system digital workers. This is the first time a major bank has publicly set a timeline for long-running autonomous AI, a signal that matters well beyond the technology itself.
From two minutes to two hours — what exactly changed?
Current AI agents typically run for just two to three minutes, completing a single task before stopping and waiting for human instruction.
The version JPMorgan is about to deploy can run autonomously for one to two hours, managing full workflows across multiple steps and different software systems.
This means → AI shifts from "a tool that does one thing when told" to "a digital worker that takes a goal and runs the entire process itself." Humans set the objective at the start and review the output at the end.
Why call it a "team manager" and not just a faster tool?
Chief Analytics Officer Derek Waldron compared the new agents to "team managers rather than individual workers" — capable of breaking down complex problems, assigning sub-tasks, and coordinating across multiple software systems.
The technical foundation behind this leap: maturing capabilities in code generation, browser control, and desktop software interaction.
In plain terms = the old AI was an intern who could only do the one thing you asked. The new version is a project manager — you hand it a big goal, and it breaks it down, delegates, and tracks progress on its own.
Waldron expects autonomous run times to extend further — from hours to days and eventually weeks.
Is it already making money — what do the private-banking numbers say?
JPMorgan has deployed AI in its private banking division, where it scans market movements, client portfolios, and research reports overnight, freeing bankers to focus on client engagement.
Waldron disclosed that these tools have driven a 20% increase in gross sales for the unit, and he expects they could eventually help a single banker expand client coverage by up to 50%.
This means → AI is not running demos in a lab — it is already pulling in revenue in a live business line. That turns the deployment of long-running agents from a technology experiment into a strategic bet backed by real financial returns.
Will there be mass layoffs — how does JPMorgan frame AI?
CEO Jamie Dimon has stated openly that some roles will be replaced by AI, and the firm is preparing to retrain and redeploy affected employees.
But Waldron's comments reveal a deeper strategic shift: the firm's framing of AI is moving from cost-cutting tool to revenue-expansion engine.
Waldron's words: "Winning the AI race isn't about cutting the most jobs — it's about building sustainable competitive advantage."
This reflects a narrative shift at the top of Wall Street: the emphasis is no longer on "how many people can we replace" but on "how much more can we earn."
What does this mean for software companies — why are moats narrowing?
Waldron said JPMorgan is increasingly evaluating its in-house build capabilities rather than relying on outside vendors.
His blunt assessment: "The moats of certain types of software companies have clearly narrowed compared to the past."
This means → when a major enterprise commands nearly $20 billion in annual technology spending, and AI dramatically lowers the barrier to building in-house, the subscription-revenue model of traditional enterprise software vendors faces a real threat.
In plain terms = banks used to buy software because they couldn't build it themselves. Now AI makes "build it yourself" fast and cheap — and the software vendors' biggest customers are turning into their biggest competitors.
Content is for reference only, not financial advice.