AI Becomes Bank Infrastructure
- Large banks are embedding AI across workflows rather than treating it as standalone experiments. - A survey notes JPMorgan, Goldman, Citi and BofA are investing billions to apply AI in coding, research and document handling. - That shift is pushing roles toward tool adoption, model governance and operational analytics (businessinsider.com).
Big banks are rebuilding work around AI, turning it into core plumbing for coding, research, customer service and document review. (businessinsider.com) At JPMorgan Chase, executives told investors on May 20, 2024 that the bank had already tied artificial intelligence use cases to an estimated $1 billion to $1.5 billion in value, spanning personalization, trading, fraud management, credit decisions and operations. The same investor-day deck said modern infrastructure and the bank’s data position it to push AI into software engineering, customer support and front-office productivity. (jpmorganchase.com) Goldman Sachs said in its 2024 annual report that it had already deployed a developer copilot coding assistant and a natural-language GS AI assistant, and that it planned to increase use of those tools in day-to-day workflows throughout 2025. CNBC reported on July 11, 2025 that Goldman had also begun piloting autonomous coding software, a step beyond chatbots that summarize text or draft emails. (goldmansachs.com) (cnbc.com) Citi told employees on June 23, 2025 that it had launched Citi AI to about 150,000 colleagues and said thousands were already using the tools daily. The bank said Citi Stylus, its document-intelligence system, had added real-time chat, a browser plug-in and Microsoft Teams integration as it pushed AI into routine internal work. (citigroup.com) Bank of America said on April 15, 2025 that more than 90% of employees were using Erica for Employees, its internal assistant, and that the tool had cut calls to the information-technology service desk by more than 50%. The bank also said its Merrill and Private Bank teams logged more than 23 million interactions in 2024 with internal AI search tools built on the same technology. (bankofamerica.com) The shift is less about one chatbot and more about where the software sits. These banks are placing AI inside the systems employees already use, which turns it from a pilot project into operating infrastructure. (jpmorganchase.com) (citigroup.com) That changes the jobs around it. Citi said it created AI Champions and AI Accelerator programs with more than 2,000 employees to drive adoption and feedback, while Bank of America said its approach includes human oversight, transparency and accountability for outcomes. (citigroup.com) (bankofamerica.com) Banks have used older forms of AI for years in fraud checks, risk models and customer targeting. What changed after the generative AI boom was the range of tasks machines could handle, especially reading documents, answering questions in plain English and helping engineers write code. (jpmorganchase.com) (goldmansachs.com) The constraints are also familiar to banks: bad outputs, privacy risks, model drift and compliance failures. That is why the public messaging from Citi and Bank of America pairs rollout numbers with controls, supervision and governance rather than promises of fully automated finance. (citigroup.com) (bankofamerica.com) The near-term result is not a bank run by bots. It is a bank where more employees are expected to use AI tools every day, and where the advantage shifts to the firms that can wire those tools into controlled, measurable workflows first. (businessinsider.com)
