Wells Fargo Cuts 8% of Staff; JPMorgan's AI Spending at $700M

The four major American banks that opened the earnings season this Tuesday (13) reported higher profits and a mixed picture of how much artificial intelligence currently impacts their costs. Wells Fargo announced that it finished the third quarter with 8% fewer employees than a year earlier. JPMorgan, which recorded a profit of $15.3 billion, a 6% increase, used the conference call to correct the publicly circulated estimate of its technology expenditures.

The number JPMorgan wanted to fix

Jeremy Barnum, CFO of JPMorgan, informed analysts that the figure of around $2 billion that circulated in the press as the bank's annual AI spending is incorrect. According to him, the actual figure is closer to $700 million per year, and even that number is already outdated. Barnum explained that the costs have spread throughout the organization to such an extent that isolating them has become difficult, and that the previous estimate probably included traditional uses of machine learning that the bank has practiced for years.

He also declined to quantify the return: according to coverage from Banking Dive, Barnum stated that the benefits of AI have yet to appear in the quarterly results.

Who is shrinking and who is hiring

Wells Fargo was the most explicit regarding its workforce. According to the earnings announcement, non-financial expenses fell by 1%, driven by efficiency initiatives that include the 8% reduction in staff over 12 months. In the conference call, CEO Charlie Scharf mentioned that the workforce decreased by 2% from the previous quarter and is 30% below the peak from the second quarter of 2020, adding that the bank is "dedicating significant management time and resources to build AI capabilities." Earnings per share rose 13% to $1.87.

Scharf did not attribute the entire cut to AI, noting that the shrinkage began before generative technology reached the branches. However, CFO Mike Santomassimo had already mentioned in September, according to American media, that AI "will further reduce the workforce" as the bank automates programming, operations, and customer service.

Goldman Sachs went in the opposite direction during the quarter. The 8-K form submitted to the SEC reports a 3% increase in the number of employees compared to the end of June, primarily attributed to the seasonal hiring of new graduates. David Solomon referred to the moment as a "very strong technology supercycle," but cautioned, according to AFP, that missteps could slow it down, and said that no one knows how demand for capital and computing will evolve.

Citigroup posted a profit of $5.2 billion, with revenue of $23.8 billion, an 8% increase and its best third quarter in a decade. Jane Fraser linked expense control to structural efficiency gains through automation and AI.

The counterpoint

A Bloomberg Opinion column published the same day argues that bank CEOs are overly optimistic about the AI cycle. The article points out that trading activity and investment banking commissions have slowed following a very strong second quarter, while executives describe the demand for capital to finance AI infrastructure as "insatiable." Thus, there are two different risks in the same statement: internal use of the technology, which so far costs less than previously thought, and credit and market exposure to the data center boom, which is significantly greater.

Where the effect is visible

In the U.S., the most visible effect is on the workforce of Wells Fargo's branches and operations, which has 30% fewer employees than in 2020. In India, the same trend coexists with hiring: JPMorgan plans to recruit around 1,000 technology professionals for its global capabilities center in the country, according to Indian media, focusing on cloud, cybersecurity, and data pipelines for AI. The centers in Bengaluru, Hyderabad, and Mumbai are precisely handling the engineering work that supports automation done at the headquarters.

For consultancies and integrators selling to these banks, the reading is uncomfortable. If the largest American bank spends $700 million annually on AI and cannot point to the return in the quarter, the pressure for proof of productivity in service contracts is likely to increase. In Brazil and the Philippines, hubs of shared services and BPO serving foreign banks are doing exactly the type of work, operations, and customer service that Santomassimo cited as targets for automation.

The missing data is the most relevant for those who buy and sell technology: none of the four banks separated, in their reports, how much of the cost reduction came from AI and how much came from existing efficiency programs. Until this line appears, the calculations of cuts attributed to AI will continue to be based on executive statements rather than audited figures.