A senior Goldman Sachs partner warned that Wall Street’s aggressive embrace of artificial intelligence risks creating cognitive atrophy among junior financiers. Speaking on a company podcast, Chris Churchman cautioned that delegating critical reasoning to algorithms threatens the traditional apprenticeship model that trains the industry’s future leaders.
The Apprenticeship Model Under Threat from Wall Street’s AI Push
Financial institutions are pouring capital into artificial intelligence to automate routine tasks, draft pitch decks, and build financial models. Yet the head of Goldman Sachs’ Marquee digital platform for institutional clients argues that rushing to eliminate grunt work carries a hidden, long-term cost. While automation drives efficiency and short-term profitability, it risks stripping away the foundational problem-solving tasks that turn junior analysts into seasoned decision-makers. The comments came during the latest episode of the firm’s "Exchanges" podcast, according to a transcript provided exclusively to CNBC.
Banking has long relied on a rigorous apprenticeship system where junior employees spend years working through complex calculations by hand, spotting patterns in raw data, and developing an intuitive feel for market dynamics. That hands-on experience builds the pattern recognition required for high-stakes finance. When algorithms handle the heavy lifting, firms risk producing a generation of workers who know how to prompt a system but cannot reason from first principles.
"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,"
Chris Churchman, partner leading Goldman Sachs’ digital platform for institutional clients, Marquee
Weighing Automation Efficiency Against Tacit Knowledge
The internal debate at Goldman Sachs cuts against a broader Wall Street gold rush where competitors tout rapid productivity gains and lean operations. Executives across knowledge industries grapple with similar tensions, questioning whether tools designed to speed up routine assignments ultimately prevent workers from developing deep expertise. Goldman has spent the past two years integrating AI across its operations, from algorithmic trading to client service chatbots. CEO David Solomon has repeatedly emphasized AI as central to the bank’s competitive strategy, while the firm’s technology division has been hiring machine learning engineers at a breakneck pace.
Historical parallels exist within the financial sector. When electronic trading replaced open-outcry floor trading, the industry shifted fundamentally. The current shift differs in both breadth and velocity, with large language models automating cognitive workflows as firms pour billions into AI tools. Last year, CNBC reported that Wall Street firms were examining ways of using AI to lower the ratio of junior bankers to senior employees.
To prevent the erosion of institutional knowledge, the co-chair of Goldman’s Global Banking and Markets AI working group emphasizes that internal systems must be built so employees retain control over high-uncertainty decisions. Even as the bank expands automation for tasks like client pricing requests, leadership acknowledges that firms must intentionally preserve the tacit knowledge that is never written down in training manuals. As Churchman, who ran currency trading at UBS before joining Goldman in 2021, noted, you learn by doing. Goldman needs to make sure it does not lose that tacit and intuitive knowledge that some of its best people have today, and to ensure the next generation has it too.
Auditing the Limits of Financial Language Models
Beyond staff development, technical implementation presents its own hurdles. In high finance, the tolerance for error is exceptionally narrow, requiring platform outputs to be fully auditable and strictly factual. Testing enterprise tools against rigorous standards often reveals the limitations of underlying language models. The toughest challenge, from a technical standpoint, is in ensuring that AI answers are 100% factual and can be audited.
During the development of the Marquee AI platform, which is used by hedge funds and other institutional clients to access Goldman’s market data, research, risk analytics and trade execution services, technical teams subjected the software to intense scrutiny. The Goldman Sachs partner noted a striking moment of machine candor when developers challenged the system hard.
"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.’"
Chris Churchman, co-chair of the Global Banking and Markets AI working group at Goldman Sachs
Keep reading