AI is taking on more analytical work at Goldman Sachs, including tasks traditionally handled by junior bankers and traders.
Chris Churchman, a Goldman partner who leads the firm’s Marquee platform, said greater reliance on AI could affect how employees develop their reasoning skills.
Much of Wall Street’s training happens on the job, with junior employees learning by working through problems alongside senior colleagues. Goldman is now weighing how automating that work could affect its next generation of bankers, including the potential for what Churchman called “cognitive atrophy.”
Goldman is confronting the issue through its own AI work.
The firm is still deciding how much to automate, where employees need to stay involved and how to meet the accuracy requirements of using AI in finance.
Why It Matters: Goldman’s experience puts workforce development directly into the enterprise AI conversation. Automating junior-level work can save time, but removing too much of it could leave future senior employees with less experience to draw from when they need to oversee AI or make decisions themselves.
- Automation Could Change How Junior Employees Learn: Wall Street has long relied on an apprenticeship model where junior employees learn by working alongside experienced colleagues. Handling client pricing requests helps junior traders understand how senior traders evaluate markets and risk. Automating those requests could reduce the firsthand experience future traders get early in their careers.
- Experience Still Needs To Be Passed Down: AI could also change how knowledge moves from senior employees to the next generation. Some of the lessons experienced bankers rely on were learned over years of working through real situations with colleagues. As more of that work is automated, Goldman is considering how newer employees will gain the same level of understanding.
- Employees Still Need To Make Decisions: High-stakes or uncertain decisions still require employees to make the final call. If the job becomes primarily reviewing and approving a model’s recommendation, employees may have fewer opportunities to develop the reasoning needed to reach those conclusions independently.
- Accuracy Remains A Challenge For Marquee AI: Goldman is bringing AI capabilities to Marquee, its platform for institutional clients to access market data, research, risk analytics and trade execution services. The AI platform is currently limited to Goldman employees. Making its answers factual and auditable has been one of the toughest technical challenges because financial applications leave little room for incorrect information.
- Convincing Answers Still Need Scrutiny: During development, Goldman pushed the system on the quality of its responses. The model described itself as “better at sounding thorough than being thorough.” The response captures the accuracy problem Goldman is working through and the need for employees who can recognize when an AI-generated answer does not hold up.
Goldman Sachs partner warned that AI is putting bankers’ reasoning skills at risk – Quartz
