JPMorgan Sees 10–30% Productivity Gains From Early AI Tools
JPMorgan global chief information officer Lori Beer told Bloomberg that the bank is already seeing 10% to 30% productivity gains from early AI tools in its technology organization, with agentic systems likely to expand the opportunity. She framed AI less as a headcount-reduction program than as a way to increase capacity for product and engineering work, while warning that the same tools raise cybersecurity risks and require tighter controls, flexible vendor choices, and leadership capable of managing through uncertainty.
Productivity gains are visible, but not as a headcount target
Lori Beer said JPMorgan is already seeing 10% to 30% productivity improvement in its technology organization from the first generation of AI tools, and expects agentic systems to increase the opportunity. Her account of enterprise AI adoption, framed on screen by Bloomberg as “enterprise AI adoption quickens on Wall Street,” is that the pace of change has accelerated sharply in the past six months: more models are being created, and JPMorgan is finding more ways to apply them.
That productivity claim came in response to Lisa Abramowicz asking about anxiety over mass job cuts, including a report she cited about Standard Chartered’s CEO saying he would cut 50% of support staff and replace lower-value human capital with AI tools and AI investment. Beer did not answer with a JPMorgan headcount target. She framed the gains less as a job-cut plan than as a way to expand capacity while demand for technology work remains high.
In Beer’s telling, JPMorgan wants to create new products, services, and customer experiences, and the bank is using AI to identify new products to offer. The effect she emphasized was getting “a lot more done” than the bank could historically.
That does not make the change benign or easy to manage. Beer also said AI is increasing cybersecurity risk, which means productivity gains come with a parallel obligation to invest in protection for customers and clients. Agentic AI may expand automation and software-development capacity, but for a bank, those gains sit inside a trust and control environment rather than outside it.
AI is both a cyber risk and a defense tool
Beer’s security argument has two sides. Asked directly by Abramowicz, “How scary is Mythos?” Beer treated AI broadly as both a source of vulnerability and a tool for defense. Models can help identify vulnerabilities, which means the bank has to decide how those findings are captured, cleared through a central process, and updated safely and securely.
The AI is also creating increased cybersecurity risk.
The countervailing point is that AI can be applied inside the software-development process itself. Beer said the use of these models in engineering workflows can help produce safer and more secure software. Cyber teams can also use the same class of tools to protect systems, defend the bank, identify fraud, and find issues more quickly.
That dual-use character explains why Beer did not describe AI adoption as a race to move at maximum speed. JPMorgan has to move quickly enough to capture new capability, but the bank also has to account for the fact that the technology changes the threat environment. In her formulation, the institution must use AI to protect itself even as AI creates new protection needs.
The harder work is leading through uncertainty
Lori Beer described AI as a “transformational shift in technology” because it is being embedded horizontally across the stack. The harder problem, she said, is change management and leadership.
The uncertainty is practical. JPMorgan does not know all the answers about where the technology is going, which providers will ultimately win, or which applications will become durable. Leaders have to guide teams without pretending that the destination is fixed. They have to help software engineers understand what is coming and create an environment that can pivot as conditions change.
Beer gave a concrete example from her own management team: senior leaders spend time on weekends building their own apps with agentic coding tools. The point was not that executives should become full-time engineers. It was that leaders need direct experience with the tools in order to understand the change they are asking their organizations to absorb.
This is a moment where leaders, in addition to software engineers, truly have to understand the change.
For Beer, that understanding has to remain tied to the bank’s actual work. AI experimentation is not an end in itself. The relevant question is how JPMorgan delivers products, services, and experiences to customers and clients while the underlying technology changes quickly and unevenly.
JPMorgan wants speed without lock-in
Lori Beer said some “moonshots are becoming real,” which changes how the bank thinks about the pace of innovation. JPMorgan, in her account, has to move aggressively at the innovation front because some capabilities that were speculative are becoming usable.
But Beer repeatedly returned to the bank’s risk posture. JPMorgan is “in the business of managing risk,” and it is also a business of trust. Customers, clients, and the bank itself have to be protected every day. The balance she described is not innovation versus risk in the abstract; it is the need to adopt new capabilities fast enough to matter while preserving the controls required of a global financial institution.
One operating implication is flexibility. Beer said JPMorgan does not know “who’s going to win in the end,” so it should not lock itself into a single provider or one scenario. The bank is constantly testing and sandboxing what is on the horizon.
That same logic shapes build-versus-buy decisions. Beer said JPMorgan has spent recent years focusing its large engineering community on building what is competitively differentiating and valuable to customers and clients. AI changes the economics because software engineering is becoming faster and more cost-effective across the end-to-end product development life cycle. That can alter the tradeoffs between building internally and buying from vendors.
Beer still drew a boundary. Some enterprise software is not competitively differentiated. It helps run the bank, but it does not deliver unique products or services to customers and clients. Those systems remain candidates for buying rather than building. The distinction is not whether AI makes software easier to produce; it is whether the software creates strategic value for JPMorgan’s customers and clients.
Banking changes, but the trust obligation stays
Asked whether banking in 10 years will still be recognizable, Lori Beer said the transformation is already underway. Her answer separated what JPMorgan expects to change from what it believes must remain fixed.
The fixed obligation is trust. JPMorgan must maintain customer and client trust and serve distinct needs across more than 100 countries. AI does not remove that requirement.
What changes is how the bank serves those needs. Beer said products and services will “absolutely evolve” over the next several years. Her view is not that AI replaces banking’s core obligation, but that it changes how a global bank builds products, manages risk, protects customers, and allocates engineering effort.