Goldman Sachs partners disclose: How 46,000 Goldman Sachs employees use AI assistants, the "biggest risk" is "over-reliance"

After generative AI has deeply integrated into the daily operations of the banking industry, Goldman Sachs partner Kerry Blum stated that while AI assistants significantly enhance efficiency, "the biggest risk is employees' over-reliance on them." On September 14, it was reported that Goldman Sachs fully launched the generative AI assistant platform, Goldman Sachs AI Assistant, to approximately 46,000 employees in June this year, which has been deeply integrated into the daily operations of banking. Partner Kerry Blum mentioned that she uses AI tools daily to handle up to 10 tasks. According to her, AI assistants have improved work efficiency in four key areas: quickly answering complex technical questions, summarizing dense document points, editing and refining written work, and brainstorming. She estimates that this technology saves her several hours of work time each week. However, Blum emphasized that bankers must recognize that AI "is a tool, not a source of truth." She stated: "The most important limitation of AI tools may be the risk of over-reliance. We must acknowledge that it is a tool, not a source of truth." Four Application Scenarios of AI Assistants Blum detailed the specific applications of AI assistants in daily work. When she encountered writer's block while communicating new projects to employees, she decided to "brainstorm with the AI assistant," which accelerated and improved her work. In a specific case, she uploaded a detailed presentation about Goldman Sachs' structured products business and asked the AI assistant to draft a three-minute introduction for an audience with varying levels of business familiarity. "It identified six key points that needed to be addressed and provided suggested comments and timing, which I used as a starting point." The AI assistant also plays an important role in document processing. The tool can shorten the drafting of IPO documents, which would normally take months, to just a few minutes, or quickly outline multi-year investment plans. Goldman Sachs has also deployed a new AI translation tool to help employees quickly translate complex financial terms and provide information in the primary language of clients. The report noted that the time saved allows Blum to devote more energy to interactions with colleagues and clients. "I can write down the content or ideas I want, and then let the AI tool finish the final work, which is very efficient for me, allowing me to move on to the next topic or spend more time interacting with our personnel." "Over-reliance" Becomes the Biggest Hidden Danger Despite the significant efficiency gains brought by AI technology, Blum emphasized that its limitations cannot be ignored. She believes that the way problems are solved is largely influenced by personal experience and a long-accumulated knowledge base, which AI tools cannot fully replace. Blum stated: "AI tools may provide a foundation, but I need to layer on the unique things that happen in any specific situation so that I can ensure I get the best answer. The tool may give a decent answer, or it may even provide a good answer, but now I have to decide if this is the right answer for my client?" She emphasized that the ultimate responsibility for decision-making still lies with the employees. The banking industry, as a demanding customer service sector, charges millions of dollars in fees, and AI tools may lack the crucial personal and nuanced judgment. Meanwhile, regulatory factors are also influencing the speed of AI adoption. Tomasz Noetzel, the author of the research report, pointed out that banks "handle vast amounts of data, which is why AI is so useful," but "some of that data may be leaked or hacked, making it both a governance risk and a regulatory compliance risk." According to a banking survey released by Bloomberg last month, 70% of the surveyed banks indicated that generative AI will be widely used or critical to their business within the next two years, while the current percentage is only 24%. Bloomberg Intelligence estimates that global banks could cut up to 200,000 jobs due to AI automation in the next 3-5 years. Noetzel's research shows that opinions in the industry regarding the impact of AI vary: some respondents estimate that their banks will lose more than 10% of employees, while others indicate that they will increase their workforce by the same margin. "This may involve eliminating some jobs, such as back-office roles, but employees can accept skill upgrades and be deployed to different roles."