Goldman Sachs AI reasoning warning: partner flags ‘cognitive atrophy’ risk for next generation of bankers

A Goldman Sachs partner has issued a Goldman Sachs AI reasoning warning, cautioning that the rapid spread of artificial intelligence (AI) across Wall Street could erode the analytical thinking skills that turn junior employees into seasoned financial professionals. Chris Churchman, who leads Goldman’s Marquee digital platform for institutional clients, made the remarks during the firm’s ‘Exchanges’ podcast, a transcript of which was provided exclusively to CNBC.

Marquee is used by hedge funds and other institutional clients to access Goldman’s market data, research, risk analytics and trade execution services. The platform’s AI capabilities are currently available to Goldman employees only.

The Goldman Sachs AI reasoning warning: ‘cognitive atrophy’ and 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,’ Churchman said. The concern is not abstract. Just as GPS navigation reduced many people’s ability to read a map, and search engines diminished the need to memorise facts, Churchman argues that delegating analytical tasks to algorithms carries a comparable cost for those learning the craft of finance.

‘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.’

Churchman is also co-chair of Goldman’s Global Banking and Markets AI working group, giving him a direct view of how the technology is being woven into the firm’s core operations. He ran currency trading at UBS before joining Goldman in 2021.

Wall Street’s apprenticeship culture under pressure

At the heart of Churchman’s concern is the traditional Wall Street apprenticeship model, in which junior employees learn by doing: fielding client requests, making small decisions under supervision, and gradually absorbing the intuition that experienced practitioners carry. AI threatens to shortcut that process by automating precisely the routine tasks that have historically served as the training ground.

‘You learn by doing, and a lot of knowledge is tacit, it was never written down,’ Churchman 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.’

He gave a concrete example from trading: junior traders have traditionally learned their craft by handling client pricing requests under the supervision of experienced risk takers. ‘We can absolutely automate that,’ Churchman said, ‘but then do we get the senior traders that fully understand?’

The stakes extend beyond individual development. CNBC reported last year that Wall Street firms were examining ways of using AI to lower the ratio of junior bankers to senior employees, raising the prospect that AI could reduce the need for entry-level roles entirely before those who do fill them have had the chance to build deep expertise.

Churchman was clear that the solution is not to resist AI, but to design systems in which employees still make the calls in high-stakes, high-uncertainty situations rather than becoming passive operators watching algorithms decide. Even Goldman, he acknowledged, has not yet ‘figured out’ how it will manage that transition.

Accuracy presents a separate but equally serious challenge. Consumer AI chatbots routinely warn users that they can make mistakes. In high finance, that tolerance is far lower, and ensuring AI answers are fully factual and auditable has proved the hardest technical problem to solve. Churchman offered a candid account of testing the Marquee AI system under pressure. When challenged hard, he said, the software acknowledged: ‘Look, in the end, I’m better at sounding thorough than being thorough.’

That admission, Churchman suggested, is precisely why human reasoning cannot be fully handed over to the models, no matter how capable they become.