A screen shows the Goldman Sachs logo displayed on the floor of the New York Stock Exchange (NYSE) in New York City on May 7, 2025.
Brendan McDermid | Reuters
A Goldman Sachs partner leading the firm’s flagship artificial‑intelligence initiative cautioned that the expanding use of AI across Wall Street could impair the analytical abilities of the next generation of financiers.
“There is a significant risk that, in the AI era, we will outsource reasoning to models, leading to cognitive atrophy that prevents us from reasoning from first principles,” said Chris Churchman, head of Goldman’s digital platform for institutional clients, Marquee.
These remarks were made in the most recent episode of the firm’s “Exchanges” podcast, according to a transcript exclusively provided to CNBC.
Much like how modern tools have diminished navigation and memorization skills, bankers may lose analytical acuity if algorithms take over the heavy lifting, Churchman warned.
“Reasoning remains essential,” he emphasized. “One must still reason about problems, structure arguments, and now we are delegating that reasoning.”
Wall Street’s drive to embed AI throughout trading and banking may constitute a devil’s bargain: while it boosts short‑term profitability, it threatens the talent pipeline needed for the future. As AI assumes routine tasks that have traditionally taught young bankers and traders to think critically, firms risk undermining the culture that develops seasoned Wall Street professionals.
Moreover, AI could reduce demand for junior bankers. Indeed, CNBC reported that Wall Street firms are exploring ways to lower the junior‑to‑senior employee ratio through AI.
Churchman said banks must strike a balance between leveraging AI and preserving Wall Street’s apprenticeship culture, recalling his experience running currency trading at UBS before joining Goldman in 2021.
“Learning comes from experience, and much of that knowledge is tacit, not captured in writing,” he noted.
Goldman must ensure that the tacit, intuitive expertise of its top professionals is retained and passed on to the next generation, Churchman said.
For example, junior traders develop their skills by handling client pricing inquiries under the guidance of seasoned risk‑takers, Churchman explained.
“We can fully automate that process,” he said, “but then will senior traders truly understand the underlying concepts?”
Systems should be built so that employees retain decision‑making authority in high‑stakes, high‑uncertainty scenarios rather than becoming passive operators, Churchman advised.
Even Goldman, one of the world’s leading investment banks, has not yet resolved how to manage the transition it has initiated, Churchman noted, also co‑chairing the firm’s Global Banking and Markets AI working group.
Error-free?
In the same podcast interview, Churchman discussed lessons learned from deploying AI within Marquee, which provides hedge funds and other institutional clients with access to Goldman’s market data, research, risk analytics, and trade execution services.
Currently, the Marquee AI platform is limited to Goldman employees, he added.
The greatest technical challenge is guaranteeing that AI responses are 100 % accurate and auditable; unlike consumer chatbots that warn of possible errors, high‑finance tolerates little margin for mistake.
Churchman observed that, during development of the client‑facing AI platform, the system made a striking admission.
“When we pushed it hard, it was honest,” Churchman said. “It admitted that, ultimately, it is better at sounding thorough than being thorough.”