Generative AI has been surrounded by much excitement and hype lately. Many believe that large language models (LLMs) are on the verge of revolutionizing corporate America, turning businesses into one big chatbot. However, the reality is quite different. Despite the pressure from CEOs and the fear of missing out (FOMO) on generative AI, the adoption of this technology is actually moving at a slower pace than anticipated.
A recent study conducted by KPMG shed light on the cautious approach of executives towards generative AI. The survey found that a majority (60%) of respondents anticipated significant long-term impact from generative AI but admitted that they were still a year or two away from implementing their first solution. This indicates that while there is a desire to harness the power of generative AI, companies are taking their time to ensure a successful and well-managed deployment.
One notable example of this careful approach is Goldman Sachs, a leading global investment banking and securities firm. Despite being an established player in implementing AI-driven tools, Goldman Sachs has not yet put any generative AI use cases into production. Marco Argenti, the CIO of Goldman Sachs, emphasized that the company is currently focused on experimentation and has set a high bar for deployment. This cautious approach is not only due to the highly regulated nature of the company but also reflects a need for accuracy and risk management.
Goldman Sachs, like other companies, wants to feel comfortable about the accuracy of generative AI before deploying it in production. Argenti highlighted the importance of having a clear expectation of return on investment (ROI) as a prerequisite for implementation. The company has been particularly interested in the use case of software development, where they have seen significant productivity gains in their experiments. However, they are not randomly running AI models but have implemented a robust platform that ensures technical, legal, and compliance checks throughout the process.
Rather than building their own LLM from scratch, Goldman Sachs is fine-tuning existing models and incorporating retrieval-augmented generation (RAG). RAG is an AI framework that retrieves facts from an external knowledge base to provide accurate and up-to-date information. Argenti stressed the importance of combining data that Goldman Sachs has with the capabilities of RAG and fine-tuning techniques. This approach allows the company to leverage their data to its fullest potential while improving the accuracy and effectiveness of generative AI applications.
Goldman Sachs wants to be methodical and mindful in its approach to generative AI. Argenti cautioned against hyper-focusing on productivity enhancement as it may not lead to sustainable differentiation. Instead, the company aims to strike a balance between seeking ROI and investing in technologies that can potentially disrupt their industry. The focus is on practical and concrete applications of generative AI in specific use cases.
Despite their cautious approach, Goldman Sachs is not falling behind in the race towards generative AI adoption. The company recognizes the potential of this technology and has been supported by its CEO and board in their efforts. While they may not be galloping at top speed, Goldman Sachs is ensuring that it has multiple “horses in the race” to explore and deploy generative AI effectively.
The adoption of generative AI in the enterprise world is slower than anticipated. Companies like Goldman Sachs, known for their AI initiatives, are taking a careful and deliberate approach to ensure accuracy, compliance, and ROI. It is clear that a mindful and methodical deployment strategy is necessary to harness the full potential of generative AI while mitigating risks. As more organizations embark on their generative AI journeys, it is crucial to strike a balance between experimentation and caution, leading to successful implementations that drive transformation and innovation.