Goldman Sachs partner warns that “outsourcing” reasoning ability to AI may create a “cognitive crisis” for the next generation of bankers

The Zhitong Finance App learned that a Goldman Sachs partner leading the bank's core artificial intelligence project issued a warning: with the rapid penetration of AI on Wall Street, the ability of the next generation of financial practitioners to think independently is at risk of being weakened. Chris Churchman (Chris Churchman), head of the Goldman Sachs institutional customer digital platform Marquee, said, “In the age of AI, it is very likely that we will outsource all of our reasoning processes to these models, which in turn will cause cognitive decline — we will lose the ability to deduce ourselves based on first principles.”

Churchman pointed out that just as modern inventions cause people to gradually lose their ability to navigate and remember, if algorithms take over all the heavy analysis, bankers' ability to analyze and judge also faces deterioration.

“Reasoning is still critical,” he stressed. “You still need to deduce the problem and construct it as a logical argument, and we are now handing over this process to AI.”

Wall Street is fully embedding AI into transactions and banking processes. This may become a “devil's deal”: it can increase industry profits in the short term, but in the long run, it may erode the talent base needed in the future. As AI takes over many of the daily basic tasks traditionally used to train junior bankers and traders, these jobs were originally a “training ground” for them to learn to think and make decisions. Once lost, companies may disrupt their apprenticeship culture of forging newcomers into veteran Wall Street talents.

The more profound impact is that AI may even directly cut the demand for junior bankers. Last year, it was reported that Wall Street agencies are studying the use of AI to reduce the ratio of junior employees to senior staff.

Churchman believes that banks must find a balance between AI applications and traditional apprenticeship training. Prior to joining Goldman Sachs in 2021, he was responsible for foreign exchange trading at UBS (UBS).

“Knowledge is acquired through practice. Many skills are hidden knowledge and have never been clearly recorded,” he said.

Goldman Sachs must “ensure that we don't lose the kind of hidden intuitive knowledge that today's top talent has and that the next generation can also master it,” Churchman said.

As an example, he said that junior traders usually gain experience by responding to clients' requests for quotes under the guidance of senior risk decision makers.

“We can fully automate it,” he said, “but if that were the case, would we be able to develop experienced traders who really understand the full picture of the market?”

System design must ensure that employees still have the final say in high-risk, high-uncertainty decisions rather than being passive operators, Churchman emphasized.

Even as the world's top investment bank, Goldman Sachs hasn't “found the answer” — how to manage this transformation that has already begun, Churchman admits. He is also the co-chair of the bank's AI working group for the global banking and marketing department.

The “zero error” conundrum

In the interview, Churchman also shared his experience in integrating AI into the Marquee platform. Marquee provides Goldman Sachs market data, research, risk analysis and trade execution services to hedge funds and other institutional clients.

He said that currently the Marquee AI platform is only open to Goldman Sachs internal employees.

From a technical perspective, the biggest challenge is to ensure that the answers output from AI are 100% accurate and traceable. Consumer-grade AI chatbots usually alert users to possible errors, but in the high-end financial sector, there is very little room for fault tolerance.

Churchman revealed that in the process of developing the customer-facing AI platform, the software gave an alarming confession: “At least it was honest when we questioned it harshly,” Churchman said. “At the end of the day, what I'm good at is sounding very thorough rather than really thorough.”