Dimon says JPMorgan does not "uniquely benefit from AI." That is the whole problem.

JPMorgan CEO Jamie Dimon has said his bank does not "uniquely benefit from AI" because everyone is now using it, a line Business Insider places at the centre of its survey of how the largest banks are deploying the technology ^[1]. That is a candid framing from the operator of a nearly $20 billion annual technology budget ^[2], and it moves the investor question from whether a bank has adopted AI to whether it can show a benefit a competitor cannot buy off the shelf.

The pressure is already visible in earnings calls. Analysts continue to press bank executives on returns and safety as concerns build about whether the scale of AI spending is justified, according to Business Insider ^[3]. Dimon's answer so far is volume: almost 1,000 use cases across fraud protection, marketing and note-taking, disclosed on the second-quarter earnings call ^[4], and a proprietary generative AI platform rolled out to more than 200,000 employees ^[5]. He has also said the bank's $2 billion AI investment has already matched its cost in savings ^[6] - roughly a tenth of the technology budget, on his own numbers ^[7].

The gradations between firms are mostly about money and measurement. Goldman Sachs put $6 billion behind technology this year ^[8], less than a third of JPMorgan's budget ^[9], and CEO David Solomon said in October he would like at least $8 billion but "I can't afford it because I've got to deliver returns" ^[10]. That is a stated 33 percent shortfall against his own preferred spend ^[11]. Citigroup's technology chief Tim Ryan oversees a $12 billion budget ^[12], double Goldman's ^[13].

Measurement is where the strategies actually diverge. JPMorgan tracks GitHub Copilot use on a dashboard that sorts developers into "light," "heavy," or "non" users ^[14], and engineers are now expected to "drive excellence" by adopting AI under updated objectives posted on the company intranet ^[15]. Goldman CIO Marco Argenti told Business Insider he is more interested in team velocity than individual usage ^[16]. One approach produces compliance metrics; the other tries to produce output metrics, and only the second is defensible to an analyst.

The structural moves matter more than the tool rollouts. JPMorgan reorganised its commercial and investment bank in February to "maximise the impact of AI," with each major business reporting to new COO Guy Halamish ^[17], and in July restructured its firmwide chief data and analytics office as AI chief Teresa Heitsenrether retired after four decades, shifting from AI infrastructure to business initiatives ^[18]. Its asset management arm is dropping external proxy advisors in the US in favour of an in-house platform called Proxy IQ ^[19] - a case where AI cuts vendor spend rather than headcount. Goldman's October memo on the third iteration of OneGS said AI would drive efficiency, slow hiring and produce a "limited reduction" of roles ^[20]. Goldman is also working with Anthropic on agents for trade and transaction accounting and client onboarding, Argenti told CNBC in February ^[21]. Citi has trained 4,000 employees as AI stewards and Ryan says firms need a mix of "metrics and pride" ^[22].

Watch whether the $2 billion-matched-in-savings claim ever appears as a line item rather than a talking point, whether Goldman's "limited reduction" stays limited as OneGS matures, and whether JPMorgan's reorganised data office starts attributing AI gains to specific business P&Ls. Until one of those happens, Dimon's concession stands: the spend is mandatory and the edge is unproven.