Codex for every role, tool, and workflow

Codex started life as OpenAI's coding assistant, the tool built to write and review software alongside developers. Now the company wants it in rooms where nobody touches a terminal. The latest rollout adds plugins, dedicated sites, and annotation features aimed squarely at analysts, marketers, designers, and investors — people who've never needed to know what a pull request is.

The framing here matters more than any single feature. OpenAI isn't just bolting a chat window onto spreadsheets. It's positioning Codex as a general-purpose work layer, something that slots into whatever tool a team already uses rather than forcing everyone into a coding-first mental model. Annotations let non-technical users mark up and refine outputs directly, which sounds small but removes a real friction point: previously, getting Codex to adjust something meant knowing enough to describe the change in developer terms.

This is very much in line with where OpenAI has been steering its enterprise pitch all year — less "look at this impressive coding demo," more "here's how your whole company touches the same underlying model." Marketers get tools shaped like marketing tools. Investors get something that looks like their existing research process, not a GitHub clone with extra steps.

What's notably absent from the announcement is much technical detail — no benchmark numbers, no specifics on model versions powering these new surfaces. That's consistent with how OpenAI tends to talk about product expansions versus model releases: this is a distribution move, not a capability one. The interesting question isn't whether Codex can technically do marketing work or investment analysis. It's whether OpenAI can convince non-engineers that a tool named after code is actually for them.