OpenAI has embedded Codex-powered agents directly into ChatGPT workspaces, giving teams the ability to deploy autonomous task runners without leaving the chat interface. The debut use case is a third-party risk management agent that screens vendors across three distinct risk categories: sanctions, financial exposure, and reputational risk.
The walkthrough is worth watching because it shows the full pipeline, not just the output. The agent ingests vendor data, runs parallel checks, and synthesizes findings into a structured report, all inside a shared team environment. That last detail matters: this is built for collaborative use, not solo power users.
The real question the video raises is how far Codex-backed agents can go inside enterprise workflows before hitting accuracy or compliance walls. OpenAI is clearly positioning workspace agents as a productivity layer for business teams, and vendor screening is a credible entry point. Watch for how the handoff between agent output and human review is handled.
[WATCH ON YOUTUBE →]