Banks are turning AI adoption into operational change

Bank of America is moving AI from pilots into daily operations

More than 200,000 Bank of America employees now use AI-enabled capabilities. They generate over 400,000 prompts each day. Those numbers matter because they show that AI has moved beyond small tests. It is becoming part of how a large bank operates.

Bank of America has approved more than 300 AI use cases. Of these, 114 involve generative AI, and 34 have been fully implemented in bank operations as of its Q2 2026 earnings call. Employees use AI for productivity tasks, software development, and more advanced agentic workflows. These workflows use AI systems that can carry out multiple steps toward a defined task rather than only produce a single response.

The applications are spreading across operations, risk, finance, technology, and client-facing teams. Relationship managers can use AI to prepare for client meetings. Developers use it to write code more efficiently. In wealth management, the bank introduced an AI-powered tool that helps financial advisers access information held in its Salesforce customer relationship management system.

Brian Moynihan, CEO of Bank of America, told investors that these systems help relationship managers prepare more thoroughly for meetings while allowing developers to code more efficiently. He said they also improve employee productivity, consistency, and client service.

Alastair Borthwick, CFO of Bank of America, pointed to strong productivity among financial advisers and said AI has been a major contributor. He also said embedding AI into workflows has reduced manual work, increased speed, and improved consistency for both employees and clients.

The important constraint is no longer access to AI models. It is converting models into approved, reliable workflows that can operate inside a regulated bank. Bank of America’s numbers make that distinction clear. More than 300 applications have approval, but only 34 are fully implemented. Moving from approval to production requires integration with existing systems, data access, controls, security, and operational ownership.

For executives, this is the useful measure of AI maturity. Employee prompt counts demonstrate adoption, but production deployments demonstrate operational change. The next question is whether those deployments produce measurable improvements in cost, cycle time, service quality, revenue, or risk.

Wells Fargo and Citi show that AI adoption is becoming an operating priority across large banks

Bank of America is not an isolated case. Wells Fargo and Citigroup are also putting AI directly into employee workflows. The common objective is practical: reduce repetitive work, make employees faster, improve client service, and shorten the time required to deliver products.

Wells Fargo launched AI Teammate. CEO Charles Scharf said the bank’s AI investments are helping improve productivity.

Citi provides a clearer measure of adoption. Nearly nine out of 10 Citi employees are using the bank’s AI tools, according to CEO Jane Fraser during Citi’s Q2 2026 earnings call. Fraser said this use is improving productivity and client experience while also supporting growth. She also said AI is helping Citi bring products to market “significantly faster.”

Citi’s approach builds on its wider technology transformation. As that program approaches completion, Fraser said business leaders can apply lessons from large-scale technology implementations to AI integration. This matters because scaling AI inside a bank is not primarily a model-selection problem. AI must connect safely to corporate data, existing applications, business processes, and governance controls before it can influence core operations.

Leadership structure is also changing around that requirement. Citi appointed Brian Saluzzo as CIO in March 2026, with scaling AI across the organization among his priorities. That places AI deployment within the bank’s broader technology operation rather than treating it as a separate experiment.

The strongest signal from Citi is its adoption rate. Getting close to 90% of employees to use AI tools indicates substantial organizational reach. But adoption alone is not a return on investment. Executives should separate usage metrics from business outcomes. The stronger measures are reductions in processing time and manual work, faster product releases, better client outcomes, controlled risk, and ultimately financial returns.

Wells Fargo and Citi therefore reinforce a broader change in banking. AI is becoming part of the operating model. Charles Scharf at Wells Fargo and Jane Fraser at Citi both connect AI with productivity. Citi goes further by linking it to client experience, growth, and faster product delivery.

The executive priority now shifts from proving that employees will use AI to proving that widespread use creates durable economic value. Citi’s nearly 90% adoption shows that distribution can reach scale. The harder task is turning that scale into measurable business performance while maintaining the controls required of a major financial institution.

BNY sees AI as a source of long-term value

BNY is taking a broader view of AI economics. CEO Robin Vince said AI is increasing employee productivity, helping the bank build better products and client experiences, and enabling new capabilities through BNY’s platforms and data. During the bank’s Q2 2026 earnings call, Vince described AI as a significant source of long-term value for clients, employees, and shareholders.

That distinction matters. Employee productivity can reduce the time or effort required for existing work. BNY is also targeting improvements in what it sells and how clients interact with the bank. Better products, new platform capabilities, and stronger use of proprietary data can influence revenue and customer retention as well as operating efficiency.

The bank’s platforms and data are particularly important. AI systems become more useful when they can work with relevant, controlled business information and fit into existing processes. For a financial institution, this also requires strong permissions, data quality, security, and oversight. AI capability alone does not create business value if employees and systems cannot safely use the information required to complete a task.

BNY’s position also gives executives a useful framework for evaluating AI investment. Cost reduction is only one measure. Product quality, employee output, client experience, speed of introducing new capabilities, and revenue opportunities can also determine returns. This is especially relevant when AI spending initially raises technology costs.

The long-term opportunity is still significant. If BNY can combine its data, platforms, and AI systems under effective controls, the technology can support both internal efficiency and client-facing innovation. The key test will be whether those capabilities generate measurable improvements in business performance over time.

JPMorgan chase expects customers to capture much of AI’s economic value

JPMorgan Chase already has nearly 1,000 live AI use cases. They cover risk, fraud, marketing, document reading, and other functions. This is one of the clearest signs that AI deployment has reached substantial operational scale at a major bank.

Yet CEO Jamie Dimon takes a more cautious position on the financial benefit to JPMorgan itself. During the bank’s Q2 2026 earnings call, he said AI should create efficiencies in parts of the company, but he does not expect it to increase margins anytime soon. AI remains expensive, and those costs rise as usage expands.

Dimon’s central argument is that customers will capture much of the benefit. “You don’t uniquely benefit from AI,” he told investors. “The ultimate beneficiary of AI will be our customers.”

This is an important economic constraint for executives. An efficiency gain does not automatically become additional profit. If competitors gain access to similar capabilities, they can also lower costs, improve products, or deliver better service. Competition can then transfer part of the economic benefit to customers rather than allowing one bank to retain all of it through higher margins.

AI also introduces significant operating costs. Models require computing capacity, supporting infrastructure, data integration, monitoring, security, and governance. At JPMorgan’s scale, deploying AI across nearly 1,000 production use cases means those costs cannot be treated as incidental. The relevant calculation is the net economic benefit after deployment and operating expenses.

That does not make AI investment less important. It changes how leaders should evaluate it. A system can be strategically valuable even when it does not immediately expand margins. Better fraud detection can reduce losses. Faster document processing can improve service and cycle times. Stronger risk capabilities can improve decision-making. Marketing applications can make customer engagement more relevant. Some of these benefits may strengthen competitiveness without appearing as a simple reduction in headcount or operating expense.

JPMorgan’s position therefore provides a useful counterweight to productivity-focused AI narratives. Nearly 1,000 live use cases demonstrate a strong commitment to the technology, while Dimon’s margin warning sets a harder standard for measuring returns. Deployment volume and employee efficiency are intermediate metrics. The larger question is who ultimately captures the economic value.

For C-suite leaders, that distinction should shape AI investment decisions. The objective should not be to maximize the number of use cases. It should be to deploy AI where the resulting improvement matters to customers, risk performance, revenue, cost, or competitive position. JPMorgan’s experience suggests that large-scale AI adoption can be strategically necessary even when the near-term margin case remains limited.

AI is becoming part of the operating model at major banks

The numbers show that AI has moved well beyond isolated experiments at several of the largest U.S. banks. More than 200,000 Bank of America employees use AI-enabled capabilities and generate over 400,000 prompts each day. Nearly nine out of 10 Citi employees use the bank’s AI tools. JPMorgan Chase has almost 1,000 live AI use cases.

The breadth of deployment is just as important. Banks are applying AI to software development, financial advice, operations, risk, fraud detection, finance, marketing, document processing, product development, and client-facing work. Wells Fargo has also launched AI Teammate as it expands employee access to the technology.

This changes the management question. The issue is no longer whether generative AI can produce useful output. Major banks have demonstrated that employees will use it and that institutions can deploy it across many functions. The harder task is converting widespread usage into controlled, measurable business outcomes.

Bank of America illustrates this gap clearly. It has more than 300 approved AI use cases, including 114 generative AI applications, but 34 are fully implemented in operations. Brian Moynihan, CEO of Bank of America, says these tools help relationship managers prepare for meetings, developers code more efficiently, and employees improve productivity and consistency. CFO Alastair Borthwick says embedded AI has reduced manual work and improved speed.

Citi is further evidence that employee adoption can reach significant scale. CEO Jane Fraser said nearly 90% of employees use Citi’s AI tools. She linked that usage to higher productivity, improved client experiences, growth, and significantly faster product delivery. The bank also appointed Brian Saluzzo as CIO in March 2026 to help scale AI across the organization.

Wells Fargo and BNY reinforce the same direction. Wells Fargo CEO Charles Scharf said the company’s AI investments are improving productivity following the launch of AI Teammate. At BNY, CEO Robin Vince takes a broader view. He says AI can increase employee productivity, improve products and client experiences, and create new capabilities through the bank’s platforms and data. Vince described AI during BNY’s Q2 2026 earnings call as a significant source of long-term value for clients, employees, and shareholders.

JPMorgan Chase introduces an important constraint to this industry trend. CEO Jamie Dimon expects AI to create efficiencies, but not necessarily higher margins in the near term. Despite having nearly 1,000 live AI use cases, JPMorgan faces the cost of scaling the technology. Dimon told investors, “You don’t uniquely benefit from AI,” and concluded that “the ultimate beneficiary of AI will be our customers.”

That point changes how executives should define AI success. Prompt volume, employee adoption, and use-case counts are useful deployment metrics. They are not measures of return on investment. Large-scale use consumes computing resources and requires integration, security, governance, monitoring, and ongoing operational support. Banks also compete against institutions adopting similar technology, which can cause productivity gains to appear as better prices, faster service, or stronger products rather than wider margins.

The next phase of banking AI will therefore depend on measurement and execution. Executives need to connect each production deployment to an economic or operational result: lower processing costs, shorter cycle times, fewer losses, higher employee output, improved customer outcomes, faster product delivery, revenue growth, or stronger risk controls. Use cases that cannot demonstrate material value will become harder to justify as AI spending grows.

The evidence from Bank of America, Citi, Wells Fargo, BNY, and JPMorgan Chase points in one direction. AI is becoming part of normal banking operations. The competitive advantage will not come from having access to AI or generating the most prompts. It will come from selecting valuable applications, deploying them safely at scale, and capturing measurable business results.

Key takeaways for decision-makers

• AI is entering daily bank operations: Bank of America has more than 200,000 employees using AI, but only 34 of its 300-plus approved use cases are fully implemented. Leaders should measure production deployment and business outcomes.
• Adoption is no longer the main barrier: Nearly 90% of Citi employees use its AI tools, while Wells Fargo is expanding AI through AI Teammate. The priority should shift from driving usage to proving gains in speed, productivity, client experience, and growth.
• AI value extends beyond cost savings: BNY expects AI to improve productivity, products, client experiences, and platform capabilities. Executives should assess AI investments against both operating efficiency and their ability to create new client and revenue value.
• Efficiency does not guarantee higher margins: JPMorgan Chase has nearly 1,000 live AI use cases, yet CEO Jamie Dimon expects customers to capture much of the benefit as AI remains costly to scale. AI business cases should account for total operating costs and competitive pressure.
• Business outcomes will separate leaders from adopters: Major banks have established that AI can reach operational scale. The next competitive test is selecting high-value use cases and linking them to measurable improvements in cost, cycle time, revenue, client outcomes, and risk.