The fastest-growing group of Codex users isn't writing code.
OpenAI shared this week that Codex now has more than 5 million weekly active users, up roughly 6x since the desktop app launched in February. The more telling number sits underneath that: knowledge workers make up about 20% of users, and they're growing more than three times faster than developers.
These are people writing reports, building spreadsheets, drafting contracts, running data analysis. Not engineers. The fastest-growing tasks on Codex right now are research and data work, not pull requests.
OpenAI is leaning into the shift. Alongside the usage numbers, the company rolled out a batch of business plugins for the ChatGPT app, hooking Codex into tools like Atlassian, GitLab, Microsoft Suite and Databricks. The pitch is that Codex can sit inside the workflows knowledge workers already live in, rather than being a separate place they have to go.
Virgin Atlantic is the case study OpenAI is putting forward. Richard Masters, the airline's VP of data and AI, said Codex is "moving into a real tool for everyone," with his team using it to turn workshop ideas into working prototypes in a couple of hours instead of weeks.
The hard numbers are coming from the engineering side, though. Neil Letchford, Virgin's VP of digital engineering, said his teams are seeing codebases shrink by 78 to 80% and refactoring jobs that used to take two weeks finish in about 30 minutes. That kind of compression is what's making non-engineers curious in the first place.
The catch is supervision. Stanford researcher Andrew Hall asked a similar agent to update one of his old research papers, and it pulled new data, ran the analyses and made the figures. When a researcher audited the output, it had a lot of errors. The work needed close expert oversight, which is exactly the kind of supervision most knowledge workers can't provide for tasks outside their own field.
OpenAI wants Codex to be the productivity tool for everyone, and the user numbers say it's getting there. The harder problem is that the people getting real value out of it right now still tend to be the kind of users who can spot when the agent is wrong. A lawyer using it for legal work, an engineer using it for code, an analyst using it for data they understand. Knowledge workers pointing Codex at something outside their expertise don't have that backstop, and the audit step is what currently makes these tools usable only to careful users. Whoever closes that gap gets to define what an AI productivity tool actually looks like next year.
