A Goldman Sachs partner leading one of the bank’s artificial intelligence initiatives has warned that widespread reliance on AI could weaken the analytical skills of the next generation of bankers if employees allow models to do too much of their reasoning.
“There’s a huge danger here that in the era of AI, we outsource our reasoning to these models, and we have cognitive atrophy that stops us being able to reason from first principles ourselves,” said Chris Churchman, who leads Goldman’s Marquee digital platform for institutional clients.
Churchman made the comments in an episode of Goldman’s “Exchanges” podcast.
His warning reflects a broader debate on Wall Street about how banks can use AI to increase productivity without eliminating the training and apprenticeship practices through which junior employees develop judgment and expertise.
Generative AI can perform tasks such as summarizing documents, analyzing data, generating code and producing draft research, potentially reducing the amount of routine work traditionally assigned to junior bankers and traders.
Churchman said that could create a problem if employees no longer learn by performing those tasks themselves.
“Reasoning is still important,” he said. “You still need to reason about [problems] and structure it into an argument, and now we’re delegating reasoning.”
Wall Street banks have accelerated investment in AI as they seek to automate parts of trading, investment banking and research. The technology could reduce costs and increase the speed at which financial institutions process information, but banks also face challenges around accuracy, oversight and the development of junior talent.
Churchman said banks need to preserve the apprenticeship model through which employees develop practical knowledge that may not be documented in manuals or databases.
“You learn by doing, and a lot of knowledge is tacit, it was never written down,” he said.
Goldman needs “to make sure we don’t lose that tacit and intuitive knowledge that some of our best people have today [and] to ensure the next generation have it too,” Churchman said.
Junior traders, for example, learn by responding to client requests for prices while working under the supervision of experienced traders who assess risk and make decisions, Churchman said.
“We can absolutely automate that,” he said, “but then do we get the senior traders that fully understand?”
The concern is that removing too much of that entry-level work could create a gap in the industry’s talent pipeline. Employees may become proficient at operating AI systems without developing the underlying judgment needed to challenge their output or make decisions when conditions fall outside the models’ assumptions.
Churchman, who previously ran currency trading at UBS and joined Goldman in 2021, said systems should be designed so employees remain responsible for decisions involving significant uncertainty and risk rather than simply accepting AI-generated recommendations.
Goldman has not yet fully resolved how it will balance those competing priorities, Churchman said. He also serves as co-chair of the firm’s Global Banking and Markets AI working group.
Churchman said accuracy is among the biggest technical challenges Goldman faces as it incorporates generative AI into its businesses.
Marquee provides institutional clients such as hedge funds with access to Goldman Sachs’ market data, research, risk analytics and trading services. Churchman said the bank’s Marquee AI platform is currently available only to Goldman employees.
In financial markets, an AI system that produces plausible but incorrect information can create significant risks. Unlike many consumer applications, financial systems often need outputs that can be verified, traced and audited.
Churchman said the challenge is making AI responses fully factual and auditable rather than merely convincing.
“When we challenged it hard, at least it was honest,” Churchman said. “It was like, ‘Look, in the end, I’m better at sounding thorough than being thorough.’”
Goldman’s experience illustrates a broader limitation of generative AI: large language models are designed to generate likely sequences of information rather than guarantee that every statement is correct. That can make them useful for tasks such as summarization and information retrieval, while requiring additional controls when their outputs are used in high-stakes financial decisions.
For banks, the challenge is therefore not simply whether AI can perform a task, but how to deploy it while retaining human oversight, institutional knowledge and accountability for decisions.
Stay ahead of the Stories shaping our world. Subscribe to Impact Newswire and join our
WhatsApp Channel for updates on global tech, business, and innovation—all in one place.
Dive deeper into the future with the Cause Effect 4.0 Podcast, where we explore the ideas, trends, and technologies driving the global AI conversation.
Got a story to share? Contact Us to reach a global audience with Impact Newswire.
Mohd Hassan has extensive experience in news gathering, editing, and writing for the newswire industry, Contact – Info@impactnews-wire.com
Discover more from Impact Newswire
Subscribe to get the latest posts sent to your email.
