The Risk of Over-Reliance on Artificial Intelligence
A Goldman Sachs partner leading one of the bank’s prominent artificial intelligence initiatives has cautioned that the widespread adoption of AI across Wall Street poses a risk of diminishing the critical thinking capabilities of the next generation of financial professionals.
Speaking during the firm’s “Exchanges” podcast, Chris Churchman, who oversees Marquee—Goldman’s digital platform for institutional clients—warned that the industry risks outsourcing its reasoning to AI models, leading to what he termed “cognitive atrophy.”
Churchman noted that the financial sector faces a complex challenge: finding the proper equilibrium between utilizing advanced AI tools and preserving the crucial, hands-on “apprenticeship culture” integral to Wall Street. He stated that Goldman Sachs itself is still determining how to manage this balance, adding that accuracy remains one of the most significant hurdles presented by generative AI technology.
Eroding Skill Sets and Tacit Knowledge
According to Churchman, the integration of AI into every facet of trading and banking processes could create a dilemma: while making the industry more profitable immediately, it might simultaneously deplete the talent pool necessary for future growth. He compared the modern situation to historical technological shifts, suggesting that if algorithms handle all complex tasks, bankers risk losing essential analytical skills.
“Reasoning is still important,” Churchman stated. “You still need to reason about [problems] and structure it into an argument, and now we’re delegating reasoning.”
He emphasized that while AI automates routine tasks, these very tasks are what traditionally teach junior bankers and traders how to think and make crucial decisions. Churchman, who previously worked in currency trading at UBS before joining Goldman in 2021, stressed the importance of preserving “tacit and intuitive knowledge”—the type of deep understanding that is acquired through practice and is rarely documented.
Churchman recommended that systems must be designed to ensure that employees remain the primary decision-makers in situations characterized by high uncertainty and high stakes, rather than simply becoming passive operators. He added that even a major institution like Goldman Sachs has not yet “figured out” how to manage this evolving transition.
Technical Hurdles and Platform Development
Churchman also provided insights into the technical aspects of implementing AI within the Marquee platform, which provides institutional clients, such as hedge funds, access to Goldman’s research, market data, risk analytics, and trade execution services. Currently, the Marquee AI platform is only accessible to Goldman employees.
He identified the primary technical challenge as guaranteeing that the AI’s output is 100% factual and verifiable. Given that the tolerance for error in high finance is extremely low, this level of accuracy is paramount. When discussing the development process, Churchman recounted a specific interaction with the software:
When we challenged it hard, at least it was honest. It was like, ‘Look, in the end, I’m better at sounding thorough than being thorough.’
