⚡ Building ChatGPT for Finance — Joseph Kim, OpenAI's Head of Financial Services

ChatGPT for Financial Services integrates enterprise-specific data sources and templates into OpenAI’s core model, addressing the limitations of generic plugin architectures. While 95% of the product relies on existing ChatGPT primitives, success hinges on domain-specific data control and addressing industry-specific workflows. Joseph Kim, Head of ChatGPT Finance, emphasizes that effective AI adoption in finance requires human-in-the-loop systems, mirroring existing hierarchical review processes in banking. Future model improvements will focus on "human taste"—aesthetic judgment and nuanced reasoning—rather than just factual accuracy. Currently, the primary bottleneck for scaling these tools is not just technical capability but the development of robust evaluation benchmarks that accurately predict professional performance. As models evolve toward 2027, they will increasingly handle complex investment theses and autonomous tasks, necessitating rigorous safety guardrails and expert-led training to ensure reliability in high-stakes environments.