ChatGPT turns due diligence into agent workflow

This is one of the clearer enterprise agent demos OpenAI has shown because third-party risk work is structured, document-heavy, and expensive to do manually, which makes it a realistic fit for automation. If workspace agents hold up outside the demo, ChatGPT starts looking like an internal workflow layer rather than just a chat UI.

• –The workflow matches the sweet spot for agents: repeatable process, clear output format, multiple systems, and a human approval step before anything sensitive happens
• –OpenAI is positioning the builder as low-friction orchestration, where teams describe a job in plain language and ChatGPT turns it into steps, tools, skills, and triggers
• –Trace visibility and preview matter here because risk teams need to inspect how a report was produced, not just accept a black-box answer
• –The third-party risk example is a strong enterprise wedge since vendor reviews are common across procurement, finance, legal, and security teams
• –The real test will be reliability on messy real-world data, since due diligence breaks down fast when sources conflict, vendors have sparse records, or reputational signals are noisy