OpenAI says Codex and ChatGPT Work now have more than 7 million weekly active users worldwide. Their use is expanding beyond software development into reports, presentations, spreadsheets, data management, and other office tasks.
More than 1 million people use Codex for work outside software development. Those activities include cleaning data, preparing recurring reports, understanding codebases, and testing internal tools.
AI Moves Into Structured Office Work
Many knowledge-work tasks follow repeatable and structured processes, creating an opening for AI assistance. AI agents can help users organize workflows, create early drafts, and refine work that must be performed repeatedly.
Gabriel Chua, an OpenAI DX Engineer, said workers do not need to be software engineers to use these tools. They do, however, need to explain the objective, provide context, set boundaries, review the output, and iterate on it.
| Work Area | Reported Use of AI |
|---|---|
| Communications teams | Creating content calendars, reusable templates, and simple social-media monitoring workflows. |
| Data analysts and researchers | Cleaning datasets, identifying patterns, and turning analysis into usable reports. |
| Product managers | Preparing daily agendas and managing administrative work. |
For communications teams, Codex can support the preparation of reusable content structures and calendars. It can also help build a simple workflow for tracking social-media posts.
Data analysts and researchers can use the tool to clean datasets and map patterns before preparing reports. The reported use case places AI in the earlier stages of analysis as well as the reporting stage.
Product managers can also use Codex as a working assistant for daily agendas and administrative tasks. The examples cited by OpenAI show the tool being applied to a range of roles rather than a single department.
Turning Dispersed Information Into Testable Prototypes
A product manager on the Codex team used AI to gather information from Slack, documents, Linear, and email. That material was then turned into a product prototype or website that the team could test directly.
The approach gives a team something concrete to review instead of relying only on static documents. A prototype can be tested, discussed, and improved as the idea develops.
Virgin Atlantic’s digital product team used ChatGPT Work, supported by Codex technology, to compare customer experiences with competing airlines. The team then organized the information into structured data for a five-year business strategy.
Chua told detikINET that work which previously required weeks could be completed in hours in some cases. The time saved can allow teams to focus more on analysis and decision-making.
Human Judgment Remains Central
OpenAI stressed that Codex is not intended to replace human expertise. Its role is to accelerate the creation of an initial draft, while deeper judgment and review remain with people.
The usefulness of an AI agent therefore depends on the quality of direction and review provided by its user. Clear goals, relevant context, defined limits, and repeated checking remain essential parts of the workflow.
