How Is AI Changing Banking This Earnings Season?

Artificial intelligence has become an unavoidable theme of this earnings season for the world's largest banks. What was once a future-facing technology initiative is now being discussed alongside revenue, margins, and capital allocation.

The latest quarterly earnings calls from JPMorgan Chase, Goldman Sachs, Citigroup, Wells Fargo, and Bank of America all dedicated significant time to AI. Executives moved beyond broad endorsements to disclose employee adoption, production deployments, investment priorities, operating costs, and efficiency gains.

The timing is notable. Just days before earnings, reports emerged that several global banks, including JPMorgan, Goldman Sachs, Citigroup, Bank of America, and Morgan Stanley, had gained access to Anthropic's latest frontier AI model, Mythos. Rather than deploying it across the workforce, the banks are reportedly evaluating how it could be used in AI-assisted cyberattacks and whether their own security systems can withstand increasingly capable models.

Separately, Barclays and Deutsche Bank have warned regulators about Mythos's cyber-offense capabilities, even without confirmed access to the model. Together, these developments highlight how frontier AI has become both a business opportunity and a new category of operational risk for financial institutions.

That same balance between adoption and caution defined this week's earnings calls.

AI Is Becoming an Earnings Metric

Banks have historically been among the most cautious adopters of generative AI due to stringent regulatory requirements regarding customer data, model risk, security, and compliance. As governance frameworks have matured alongside enterprise AI deployments, executives now appear increasingly comfortable discussing AI with investors using measurable business metrics rather than aspirational language.

The common thread across the earnings calls was clear: AI is no longer an experimental technology. It is becoming part of the operational scorecard.

JPMorgan Chairman and Chief Executive Officer Jamie Dimon said the bank has nearly 1,000 AI use cases under development, with around 50 considered strategically important across risk management, fraud detection, marketing, and document processing.

Bank of America Chairman and Chief Executive Officer Brian Moynihan reported more than 300 approved AI use cases, including 114 live generative AI applications, 34 of which are fully implemented. He added that more than 200,000 employees now generate over 400,000 AI prompts each day.

Citigroup Chief Executive Officer Jane Fraser said nearly nine out of ten employees are already using the bank's internal AI tools.

Rather than speaking broadly about AI, banks are increasingly identifying products that have reached customers.

Bank of America highlighted relationship managers using AI to prepare for client meetings, while software developers use it to accelerate coding. Citi pointed to Citi Payments Express and Citi Wealth Advisor Insights as products that reached the market faster through AI-enabled development.

Wells Fargo highlighted Advisor Gateway, a generative AI platform for financial advisors, built following roughly $1 billion in technology investment by its Wealth & Investment Management division over the past several years.

Cost and Efficiency Are Becoming the Next AI Conversation

While adoption dominated the discussion, operating costs emerged as a second major theme.

JPMorgan Chief Financial Officer Jeremy Barnum said AI token costs remained insignificant during the first half of 2026 but are expected to increase meaningfully during the second half. To control costs, the bank is increasingly routing workloads to smaller, less expensive models instead of defaulting to frontier models for every task.

Barnum cited report summarization as an example of work that does not require the most advanced models available.

The comments reflect a broader trend emerging across enterprise AI deployments. After two years of focusing on capability, organizations are increasingly optimizing for inference costs and model selection. The discussion is shifting from simply adopting AI to deciding which models to use for specific workloads.

The Workforce Conversation Is Becoming More Nuanced

Headcount was another recurring theme, although banks described its impact differently.

Wells Fargo reported its 24th consecutive quarter of workforce reductions, with headcount falling to 197,000 employees, down 15,000 from a year earlier. Chief Financial Officer Michael Santomassimo said technology and AI are helping the bank improve efficiency faster than before.

Dimon described a more targeted approach. He said JPMorgan has reduced staffing by 30% to 40% in selected business areas, while emphasizing that most affected employees have been redeployed elsewhere within the organization.

He also argued that AI's efficiency gains are unlikely to become a lasting competitive advantage because competing banks have access to the same technologies. Instead, he suggested that customers would ultimately benefit from improved services rather than banks permanently expanding their profit margins.

That perspective differs from the broader narrative that AI will simply reduce labor costs. Instead, it suggests that banking may become another industry where AI raises the competitive baseline rather than creating long-term differentiation.

Capital Spending Remains the Bigger Story

The internal AI deployments discussed during earnings are unfolding against an unprecedented wave of AI infrastructure investment.

Dimon estimated that annual industry-wide AI capital expenditure has grown from roughly $400 billion to about $700 billion and could exceed $1 trillion next year, compared with approximately $4 trillion in annual US capital expenditure.

Goldman Sachs is positioning itself to benefit from that investment cycle beyond its own internal AI adoption.

Chairman and Chief Executive Officer David Solomon said the AI infrastructure buildout is creating financing opportunities across data centers, energy, and enterprise AI, describing it as a multi-year investment cycle that is still in its early stages.

"We are in the relative early innings of a very, very significant AI buildout cycle," Solomon told analysts.

Fraser outlined a similarly long-term strategy from Citi's perspective. She said the bank intends to accelerate planned investments in AI, automation, technology, and marketing if market conditions remain favorable, describing the additional spending as "100% on the offense." Executives said some of those investments had originally been scheduled for 2027.

AI Adoption and AI Risk Are Advancing Together

Taken together, the earnings calls suggest that AI has become something banks now measure as rigorously as they manage it. Employee adoption, production deployments, infrastructure spending, operating costs, and efficiency gains are increasingly being reported alongside traditional financial metrics.

That evolution coincides with another important development. The same institutions expanding AI across software development, fraud detection, wealth management, customer service, and internal operations are also evaluating frontier models such as Anthropic's Mythos to understand the cybersecurity risks posed by increasingly capable AI systems.

For one of the world's most heavily regulated industries, those are not separate stories. They are two sides of the same transition.

The latest earnings season suggests that AI has entered a new phase in banking. Adoption is no longer measured by pilot projects or proof-of-concept deployments. It is becoming a business metric investors expect to hear about, while risk assessment is advancing in parallel as an equally important part of enterprise AI strategy.

>

>