The bank used AI to shrink. Then it lent money to build more AI. Then the price of money moved.
In Q3 2026 financial services stopped describing AI as a pilot program. The largest banks and payment networks reported it in headcount and severance, and Mastercard started testing a risk score for payments that AI agents initiate. In the same quarter, with energy prices climbing, the Fed turned to hikes, the 10-year Treasury crossed 5% and private credit defaults hit a record, while banks financed the data centers the whole AI thesis depends on. The industry ended the quarter long AI on both sides of its balance sheet.
First Financial Services edition; these scores set the baseline. The placeholder card rated Quantum 5. In-period evidence shows supervisors moving with no binding deadline for firms, so it lands at 3.
AI moved from the innovation slide to the headcount line.
The Q2 earnings calls in mid-July were the round where the largest US banks connected AI to jobs in plain words. Jamie Dimon told analysts on July 14 that JPMorgan had "discrete areas where we did reduce jobs by 30% or 40%," with most of those people offered roles elsewhere. He also cited almost 1,000 AI use cases, and the bank's adjusted expense outlook stood at about $107.5 billion.