Banks push AI enterprise‑wide

JPMorgan is no longer talking about artificial intelligence like a lab project sitting off to the side. In Jamie Dimon’s 2025 shareholder letter, he wrote that artificial intelligence will affect “every job” and “every function” at the bank. (jpmorganchase.com) That is a very different message from the old bank playbook, where new technology usually started in one back-office team and spread slowly. JPMorgan has more than 300,000 employees globally, so saying “every function” means this is being aimed at traders, call centers, compliance teams, software engineers, and risk managers at once. (jpmorganchase.com 1) (jpmorganchase.com 2) Banks are pushing this now because they sit on huge piles of text, numbers, and rules. A large language model is useful in a bank for the same reason it is useful in a law office: it can read dense documents, summarize them, draft first versions, and surface patterns for a human to check. (jpmorganchase.com) (fsisac.com) The hard part is not getting a model to answer a question in a demo. The hard part is connecting that model to real workflows, real permissions, real customer data, and the audit trails a regulated bank has to keep. (jpmorganchase.com) (fsisac.com) JPMorgan’s own security work shows what that looks like in practice. The bank built an Artificial Intelligence Threat Modeling Co-Pilot to help architects and developers identify cyber threats and recommended controls early in the software-building process, instead of trying to patch problems after launch. (jpmorganchase.com) That security angle is why cross-industry groups are getting involved. Financial Services Information Sharing and Analysis Center, the sector’s main cyber-sharing network, said in February 2024 that it had released six white papers on artificial intelligence risks for financial firms because the technology creates both new use cases and new attack surfaces. (fsisac.com 1) (fsisac.com 2) Project Glasswing fits into that same problem set. It is part of a wider effort to treat artificial intelligence risk less like a gadget problem and more like a system problem, where banks, security teams, and infrastructure providers need common ways to test what happens when models are wrong, manipulated, or plugged into sensitive systems. (fsisac.com) (jpmorganchase.com) The hiring signal is just as important as the technology signal. If a bank is moving from pilots to production, it needs fewer people who can merely prompt a chatbot and more people who can wire models into databases, approval chains, cybersecurity controls, and day-to-day operations. (jpmorganchase.com 1) (jpmorganchase.com 2) JPMorgan is also widening the funnel for that talent. Its Autism at Work program began as a pilot in Delaware in 2015 and has expanded to more than 40 roles across nine countries, while its BeST program is designed to create career paths for neurodivergent and disabled employees. (jpmorganchase.com) (jpmorganchase.com) The reason those two moves belong in the same story is that enterprise artificial intelligence creates a lot of work that rewards unusual kinds of focus. Model testing, exception handling, controls design, and process mapping are the kinds of jobs where banks want people who can spot patterns, inconsistencies, and edge cases before software touches money. (jpmorganchase.com) (jpmorganchase.com) So the shift is not just “banks are using artificial intelligence.” The shift is that one of the world’s biggest banks is reorganizing around the idea that artificial intelligence will sit inside ordinary work, and that forces it to rebuild security rules, software pipelines, and hiring at the same time. (jpmorganchase.com) (jpmorganchase.com)