Most AI tools today work in quick bursts. You ask, they answer, they stop.
JPMorgan wants AI that keeps going. It is building agents that run on their own for an hour or two.
The goal is for one agent to finish a whole job. Not just a single step.
Derek Waldron runs analytics at the bank. He calls this the era of long-running agents.
In plain terms, the software is growing up. It is shifting from a helper into a worker.
Waldron has a name for the key trait: intellectual coherence. It means how long an AI can stay on task before a human steps in.
Better reasoning is what makes it possible. The model can plan, not just answer.
He compares it to a manager. A good manager breaks a big job into parts and hands them out.
The AI is starting to work the same way. It acts less like a calculator and more like a team lead.
The leap is less about raw smarts and more about stamina. Can the AI hold a plan together for hours?
What changed? AI can now write code, click around the web, and use desktop apps.
Viral tools over the past year showed it works. The tech is close to ready for big firms.
Today's agents run an hour or two. Waldron says they could one day last for days, then weeks.
That only happens once the safety questions are solved.
We follow how AI is reshaping real companies in Market Briefs - five minutes a day, with a free investing masterclass when you sign up.
This is not just a someday promise. The tools are already working.
JPMorgan says they have lifted sales in its private bank by 20%. The AI scans markets and research overnight.
That frees bankers to spend the day with clients. The bank thinks each banker could one day handle 50% more.
Overnight, the AI reads the news and flags client moves. By morning, the banker's prep is done.
That would be a big jump in output. It is also why the bank sees AI as growth, not just cost.
For years, firms saw AI mainly as a way to cut staff. Waldron says the real win is a lasting edge.
CEO Jamie Dimon has been honest about the downside. Some jobs will go away.
The bank says it plans to retrain those workers for new roles.
There is a warning here for software firms. Waldron said their moat is shrinking.
A moat is the edge that keeps customers from leaving. For some software makers, that edge is fading.
Build-versus-buy is the quiet shift here. More in-house tools mean fewer checks to outside vendors.
JPMorgan is the biggest US bank by assets, and Jamie Dimon has run it since 2006. It now builds more of its own tools instead of buying them.
It has a $20 billion tech budget to do it. So this is a real threat to its vendors.
JPMorgan is not just using AI. It is deciding which software firms it still needs.
Want to see where AI money is heading next? Read Market Briefs every morning and get a free 45-minute investing course when you join.