OpenAI Plans to Automate Investment Theses by 2027
OpenAI's Joseph Kim explains why financial AI is moving past junior analyst grunt work toward managing director decisions by 2027.
Building ChatGPT for Finance — Joseph Kim, OpenAI's Head of Financial Services
With swyx, Richard, Joseph Kim · Sunday, November 1, 2026
Joseph Kim, lead of ChatGPT for Financial Services at OpenAI, discusses the engineering and product decisions behind OpenAI's specialized enterprise finance tooling. He explains why bundling reliable financial data proved essential over flaky third-party connectors, how upstream training and plugin architecture enable startups to build on foundation models, and how the team is tackling human taste, review chains, and domain-specific benchmarks.
Key takeaways
- OpenAI expects financial AI to shift from entry-level workflows to associate and managing director decisions by 2027. Read more →
- OpenAI omitted the GDP eval from the GPT-6 Astra launch because the benchmark was saturated, and the untested portions were broken. Read more →
- Joseph Kim, lead of ChatGPT for Financial Services at OpenAI, builds agent safety around existing corporate review hierarchies instead of novel alignment mechanisms. Read more →
- OpenAI tested third-party Model Context Protocol (MCP) connectors from prominent data providers and uncovered error rates exceeding 40 percent. Read more →
- OpenAI does not train isolated, proprietary models for finance; it feeds financial reasoning data directly into the core foundation model run, dubbed "the big run" (such as runs like 6 Astra or 61 Astra). Read more →
- Factual hallucinations in finance are fading rapidly; the real barrier to enterprise adoption is human aesthetic judgment and subjective taste. Read more →
6 articles from this episode
OpenAI Found Financial Evals Were Broken After Astra
OpenAI found legacy financial evals were saturated and full of bad answer keys. Joseph Kim explains why they built an analyst eval.
Joseph Kim on Why Financial AI Must Copy Banking Chains
Joseph Kim explains how OpenAI secures financial AI agents by borrowing Google readability rules and investment banking review chains.
Why OpenAI Stopped Trusting Financial MCPs
OpenAI tested third-party financial MCPs and found 40% error rates. Here is why Joseph Kim chose to bundle data directly instead.
OpenAI Trains Finance Upstream: Why Startups Own the Last 20 Percent
OpenAI puts financial reasoning into core models. Joseph Kim explains why 95% of vertical tooling uses basic primitives and where startups win.
OpenAI's Joseph Kim on Teaching Taste to Finance Models
Joseph Kim explains why factual hallucination is fading in finance AI, leaving human aesthetic judgment as the real engineering hurdle.
More Latent Space episodes
- How Anthropic ships the #1 LLM in the World — with Katelyn Lesse and Angela JiangNov 1 · 4 articles
- An Eval for AI Scientists That Knows the Answer Key - Ian Quigley, Leash BioNov 1 · 4 articles
- Why Agents Love Supabase — with Paul Copplestone and Ant WilsonNov 1 · 5 articles
- The Recommendation Engine for Commodity Traders — Anekha Sokhal, MoshiOct 25 · 4 articles