Stories you may like
Wall Street’s AI gains are here — banks plan for fewer people
By December 2025, AI adoption on Wall Street had moved past experiments inside large US banks and into everyday operations. Speaking at a Goldman Sachs financial-services conference in New York on 9 December, bank executives described AI—particularly generative AI—as an operational upgrade already lifting productivity across engineering, operations, and customer service.
The same discussion also surfaced a harder reality. If banks can produce more with the same teams, some roles may no longer be required at current levels once demand stabilises.
How Wall Street banks say AI is delivering results today
JPMorgan: operational gains begin to compound
Marianne Lake, chief executive of consumer and community banking at JPMorgan, said productivity in areas using AI has risen to around 6%, up from roughly 3% before deployment. She added that operations roles could eventually see productivity gains of 40% to 50% as AI becomes part of routine work.
Those gains rest on deliberate choices rather than broad experimentation. JPMorgan has focused on secure internal access to large language models, targeted changes to workflows, and tight controls on how data is used. The bank has described its internal “LLM Suite” as a controlled setting where staff can draft and summarise content using large language models.
Wells Fargo: output rising ahead of staffing changes
Wells Fargo CEO Charlie Scharf said the bank has not reduced headcount because of AI so far, but noted that it is “getting a lot more done.” He said management expects to find areas where fewer people are needed as productivity improves.
In comments reported the same day, Scharf said the bank’s internal budgets already point to a smaller workforce by 2026, even before factoring in AI’s full impact. He also flagged higher severance costs, suggesting preparations for future adjustments are under way.
PNC: AI speeds up a long-running shift
PNC CEO Bill Demchak positioned AI as an accelerator rather than a new direction. He said the bank’s headcount has stayed largely flat for about a decade, even as the business expanded. That stability, he said, came from automation and branch optimisation, with AI likely to push the trend further.
Citigroup: gains in software and customer support
Citi’s incoming CFO Gonzalo Luchetti said the bank has recorded a 9% productivity improvement in software development. That mirrors a broader pattern across large firms adopting AI copilots to support coding work.
He also pointed to two customer service areas where AI is helping: improving self-service so fewer calls reach agents, and supporting agents in real time when customers do need to speak with a person.
Goldman Sachs: workflow changes paired with hiring restraint
According to Reuters, Goldman Sachs’ internal “OneGS 3.0” programme has focused on using AI to improve sales processes and client onboarding. It has also targeted process-heavy functions such as lending workflows, regulatory reporting, and vendor management.
These changes are unfolding alongside job cuts and a slower pace of hiring, linking workflow redesign directly to staffing decisions.
Where Wall Street banks see the earliest AI productivity gains
Across banks, the clearest gains are showing up in work that relies heavily on documents, follows repeatable steps, and operates within defined rules. Generative AI can shorten the time needed to search for information, summarise material, draft content, and move work through approval chains—especially when paired with structured processes and human checks.
Common areas seeing early impact include:
- Operations: drafting responses, summarising cases, and resolving exceptions more quickly
- Software development: generating code, writing tests, refactoring, and producing documentation
- Customer service: stronger self-service combined with real-time support for agents
- Sales support and onboarding: pulling data from documents, filling forms, and speeding up client setup
- Regulatory reporting: assembling narratives and evidence faster, under strict review and controls
Why governance shapes the pace of adoption
For banks, enthusiasm is not the main constraint. Control is. US regulators have long required strong oversight of models, and those expectations extend to AI systems. Guidance such as the Federal Reserve and OCC’s SR 11-7 sets standards for model development, validation, and ongoing review. A 2025 report from the US Government Accountability Office noted that existing model risk management principles already apply to AI, including testing and independent oversight.
In practice, this pushes banks toward designs that can be examined and traced. AI use is often limited in how independently it can act. Prompts and outputs are logged, performance is monitored for drift, and humans remain responsible for high-impact decisions such as lending, dispute handling, and official reporting.
Productivity rises, but employment questions remain
The comments from bank leaders point to a phased shift. The first phase looks like stable headcount paired with higher output as AI tools spread across teams. The second phase begins once those gains become consistent enough to influence staffing plans, through attrition, role changes, or targeted cuts.
Signals from Wells Fargo around 2026 headcount planning and severance costs suggest some banks are approaching that second stage.
At a broader level, institutions such as the International Monetary Fund have warned that AI could affect a large share of jobs worldwide, with different mixes of automation and augmentation depending on role and region. The World Economic Forum’s Future of Jobs Report 2025 also projects substantial job movement as companies adopt AI and adjust skill needs.
What AI means for Wall Street bank strategy beyond 2025
Banks that gain the most from AI are likely to focus on three areas at once: redesigning workflows rather than layering on chat tools, building strong data foundations, and putting governance in place that supports speed without eroding trust.
User's Comments
No comments there.