The Goldman Sachs corporate logo is displayed on a screen on the floor of the New York Stock Exchange (NYSE) on May 7, 2025 in New York City, USA.
Brendan McDiarmid | Reuters
a goldman sachs A partner leading one of the bank’s flagship artificial intelligence projects warned that the pervasiveness of AI across Wall Street risks inhibiting the thinking ability of the next generation of financiers.
“In the age of AI, there is a huge risk that we will outsource our reasoning to these models and our cognitive atrophy will leave us unable to reason from first principles ourselves,” said Chris Churchman, head of Marquee, Goldman’s digital platform for institutional investors.
The comments were made during the latest episode of the company’s “Exchanges” podcast, according to a transcript provided exclusively to CNBC.
Just as modern inventions have caused people to lose navigation and memory skills, Bankers risk losing their analytical abilities if algorithms do all the heavy lifting, Churchman said.
“Inference is still important,” he said. “You have to reason[the problem]and frame it into an argument, and now you’re delegating the reasoning.”
Wall Street’s efforts to incorporate AI into all trading and banking processes may be a devil’s bargain. This could make the industry more profitable today while eroding the talent needed for the future. As AI takes over many of the routine tasks that traditionally taught young bankers and traders how to think and make decisions, companies risk sacrificing the culture that turns young employees into Wall Street veterans.
It might even reduce the need for junior bankers in the first place. CNBC reported last year that Wall Street firms were looking at ways to use AI to reduce the ratio of junior bankers to senior employees.
Churchman, who ran currency trading at UBS before joining Goldman in 2021, said banks need to find a balance between using AI and maintaining Wall Street’s apprentice culture.
“You learn by doing. A lot of knowledge is tacit and never documented,” he says.
“We need to make sure that we don’t lose the tacit, intuitive knowledge that our best people have today, and we need to make sure that the next generation has it too,” Mr. Churchman said.
For example, junior traders learn by responding to customer pricing requests under the supervision of experienced risk-takers, Churchman said.
“You can fully automate it,” he said. “But will it be completely understandable to advanced traders?”
Churchman said systems need to be designed to allow employees to make high-stakes decisions with high uncertainty, rather than being passive operators.
Even Goldman, one of the world’s largest investment banks, has not yet “thought” how to manage the transition it has begun, said Churchman, who also co-chairs the firm’s Global Banking Markets AI working group.
Are there any errors?
In the podcast interview, Churchman also shared lessons learned from implementing AI into Marquee, which hedge funds and other institutional investors use to access Goldman’s market data, research, risk analysis and trade execution services.
He said the Marquee AI platform is currently available only to Goldman employees.
From a technical perspective, the most difficult challenge is ensuring that the AI’s answers are 100% factual and auditable, he said. Consumer AI chatbots alert users to potential mistakes, but in high-value finance, there is less tolerance for mistakes.
Churchman said that while developing the company’s AI platform for customers, the software made some surprising confessions.
“When we tried hard, at least it was honest,” Churchman said. “I was like, ‘Look, at the end of the day, I’m better at being thorough than being thorough.'”
