Dimon says AI already cut 30 to 40 percent of jobs in some JPMorgan units
The bank posted $21.2 billion in quarterly profit, runs AI across nearly 1,000 use cases, and its CFO just became one of the first to treat token spend as a line item worth forecasting.
- Jamie Dimon said AI has already reduced jobs by 30 to 40 percent in discrete areas of JPMorgan, speaking on the bank's July 14 second-quarter earnings call; most affected employees were offered other roles.
- JPMorgan runs AI across nearly 1,000 use cases on a technology budget of almost $20 billion a year, and reported $21.2 billion in quarterly net income, up 41 percent.
- CFO Jeremy Barnum called token-related AI costs trivial today but forecast meaningful acceleration in the second half of 2026, one of the first bank CFOs to treat inference spend as a tracked expense.
The biggest bank in America put hard numbers on agent-era attrition. On JPMorgan's second-quarter earnings call on July 14, CEO Jamie Dimon said the bank has "had discrete areas where we did reduce jobs by 30% or 40%" because of AI, with most affected employees offered positions elsewhere, as Fast Company reported. The disclosure came wrapped in a blowout quarter: $21.2 billion in net income, up 41 percent year over year.
A thousand use cases and a token budget
JPMorgan says it now runs AI across nearly 1,000 use cases, from fraud protection to marketing to meeting notes, on a technology budget of almost $20 billion a year, per the call coverage. The quieter but arguably more important line came from CFO Jeremy Barnum, who described token-related AI costs as "trivial" today while forecasting "some meaningful acceleration" in the second half of 2026. A major bank CFO forecasting inference spend as an expense line is precisely the world the agent cost math predicted: volume grows faster than prices fall.
Dimon's margin realism
Dimon also punctured the fattest assumption in enterprise AI: that efficiency gains stay with the company that captures them. "You don't uniquely benefit from AI," he argued, noting that if automation accrued to the bottom line, two decades of computerization would have produced 80 percent margins. Competition passes the gains to customers. That framing lands differently at a bank that built agents good enough to beat the 60/40 portfolio and still refused to deploy them, and it sharpens the labor story running all summer, from the early-retirement wave to the 69 percent of Americans who want AI companies to hand over half their stock. The jobs are moving faster than the margins, exactly as Dimon says, and now there is an earnings call on the record proving both halves.