JPMorgan Chase has accelerated its use of artificial intelligence, expanding its LLM Suite platform to support 1,000 business applications. However, the bank’s CIO, Lori Beer, emphasized that security remains the top priority in all AI initiatives.
Security First Approach
“We have to make sure security, how we manage risk, is front and center of everything we do,” Beer stated during the Evident AI Symposium on Thursday.
Her remarks followed CEO Jamie Dimon’s warnings about rogue agents, prompting the bank to prohibit outside agents from operating within its environment. Across financial services, executives are confronting the security challenges posed by rapidly advancing AI technologies, particularly as autonomous agents become more prevalent.
Marco Argenti, CIO at Goldman Sachs, highlighted the difficulty of aligning AI agents with enterprise standards. “They’re not deterministic machines, they’re stochastic machines.”
Beer stressed the need for foundational technology controls such as identity and access management, privileged access, and environment separation to be “best in class.” She noted that moving from deterministic to non-deterministic systems introduces new risk pathways requiring updated security frameworks.
Controlled Development and Risk Management
“We are going to start with one core agent-building platform,” Beer explained, adding that containing development allows the bank to manage risk and implement controls at scale through autonomous methods. She acknowledged the complexity of dynamically implementing guardrails and real-time evaluations.
“This is one of the most important times to make sure you are appropriately balancing risk and innovation tradeoffs,” Beer said. “We want to move at speed, but sometimes we have to pull back to get controls in place. The technology is running ahead of capabilities to manage it.”
JPMorgan’s history of prioritizing security includes blocking ChatGPT in 2022 due to potential data loss risks. The bank developed its own ecosystem within a private cloud built on Microsoft Azure, enabling data scientists to access secure resources. This initiative led to the creation of the LLM Suite, now used by 200,000 employees.
Measuring Business Value
While security remains the top priority, Beer emphasized that cost management and business value delivery are equally critical. Coding assistants represent the largest portion of the bank’s token expenditure. To better measure impact, JPMorgan has shifted from tracking use cases to focusing on 50 strategic initiatives the business owns as a way to measure tangible value.
“It is incumbent upon us that the investments we make ultimately translate, not to a business case on paper, but real business growth, real business outcomes, and that’s the journey we’re on,” Beer said.
Forward-Deployed Engineers and Industry Collaboration
As AI agents evolve, Goldman Sachs partnered with Anthropic this year to launch an AI services company that embeds engineers within midsize firms to implement and scale Claude.
“Obviously, those companies are ahead of you,” Argenti noted. “To learn and compress the time of learning, having people directly from the lab working with your people is the best option to train.”
He described the engagement model as distinct from traditional consulting: “This is a specific kind of engagement whose job is to become unnecessary in the shortest time possible and move to the next difficult problem. You come in, make sure the team builds the muscle, and then you can move on your own.”
Similar FDE programs are being developed by Google, Microsoft, and Amazon Web Services, which are investing billions to embed specialized engineers within enterprises to help IT leaders overcome scaling challenges.
If IT leaders maintain awareness that AI agents may produce unexpected results, Argenti advised establishing containment measures and imposing rules on agents. He also recommended creating user tool gates before agents can execute transactions.
