Goldman Partner Warns AI Could Erode Bankers’ Reasoning
Goldman Sachs partner Chris Churchman warned heavy reliance on AI could ‘outsource our reasoning’ and cause cognitive atrophy that weakens bankers’ first-principles reasoning.
On an internal podcast, Goldman Sachs partner Chris Churchman warned heavy reliance on artificial intelligence could ‘outsource our reasoning’ and lead to ‘cognitive atrophy’ that reduces bankers’ ability to reason from first principles. Transcripts of the discussion were later made available to the press.
Churchman leads Marquee, Goldman’s digital platform for institutional clients and one of the firm’s main AI initiatives. He argued banks risk delegating too much judgment to models, which could leave employees less able to think through problems independently.
He described many trading and risk-taking skills as learned through practice and direct supervision and not fully captured in written procedures. ‘You learn by doing, and a lot of knowledge is tacit, it was never written down,’ he observed. He gave the example of junior traders refining pricing instincts by handling client requests while senior traders oversee and correct decisions. ‘We can absolutely automate that,’ he added, ‘but then do we get the senior traders that fully understand?’
Goldman is integrating AI tools into workflows to improve speed and scale. Churchman called for a balance between using automation for routine tasks and preserving the apprenticeship culture that helps staff develop judgment. He noted that if systems take over more analytic work, employees would have fewer opportunities to gain hands-on experience.
The bank plans to hire roughly 2,400 to 2,500 interns this year and expects a similar-sized group of permanent new hires to start in July. In separate remarks, CEO David Solomon acknowledged hiring will ‘contract a little’ over the next three years and described the change as a modest recalibration rather than a retreat from recruiting.
Churchman’s comments framed a practical question for firms adopting AI: how to use models for efficiency while ensuring staff gain the exposure needed to manage complex or novel situations models may not anticipate.








