JPMorgan Chase says it plans to deploy “long-running autonomous” AI agents later this year — systems meant to move beyond one-off prompts and instead manage multi-step workflows across different software tools. Derek Waldron, the bank’s chief analytics officer, says these agents can run for “an hour or two,” compared with the “two or three minutes” typical of earlier versions. In bank-speak, that’s less “helpful assistant” and more “digital coworker who doesn’t know when to stop.”
Waldron frames the next battleground as “intellectual coherence”: how long an AI can operate effectively before a human has to step in. He credits improvements in model reasoning — plus practical abilities like writing code, controlling web browsers, and interacting with desktop software — for making agents more like “team managers” delegating tasks than single-purpose doers. JPMorgan’s planned deployment suggests the technology may be getting closer to clearing the security and governance hurdles that have slowed corporate adoption, though Waldron still says long-running agents “aren’t yet ready for corporate use” and expects them in 2026.
The bank is also pointing to revenue impact, not just efficiency. In private banking, JPMorgan uses AI to screen market activity, client positions, and research overnight, helping bankers focus on client interactions. Waldron says those tools have led to a 20% increase in gross sales, and that they could eventually let individual bankers expand client coverage by as much as 50%. Meanwhile, CEO Jamie Dimon has said some workers will be displaced — paired with plans to train and redeploy affected employees — while Waldron notes a more strategic shift: building more capabilities in-house and potentially squeezing software vendors as “the moat around certain types of software companies is… diminished.”
Read more at CNBC.