OpenAI adds scheduled automations to Codex

OpenAI says Codex can now run tasks automatically on schedules and triggers, turning the coding assistant from a mostly interactive tool into something that can act on its own at set times. The company announced the feature in a post on OpenAI Academy, describing it as a way to make Codex more proactive for recurring work.

Codex is OpenAI's coding assistant, designed to help with software tasks in a chat-like workflow. Until now, the model has been mainly associated with answering prompts, generating code, and helping developers work through specific requests. With automations, OpenAI is adding a layer that lets Codex show up without waiting for a fresh prompt every time.

According to OpenAI, the feature can be used for tasks such as morning briefs, weekly reviews, status updates, and data cleanup. That makes it closer to a scheduled worker than a one-off helper, which is a meaningful shift for teams that already use AI to handle repetitive software and operations work.

The basic idea is simple. A schedule or trigger tells Codex when to run, and the system uses the context it has been given to carry out the task. In practical terms, that could mean preparing a summary at the start of the day, checking on open work once a week, or handling routine maintenance steps that do not need a human to type the same request every time.

This fits a broader trend in AI tooling, where assistants are being tied into workflows instead of staying confined to chat windows. For developers and operations teams, the appeal is less about novelty and more about reducing the amount of repetitive prompting needed to keep routine jobs moving.

The mention of morning briefs and weekly reviews also points to the kind of work OpenAI expects people to automate first. These are tasks that are structured, repeat on a schedule, and usually depend on pulling together information from a known set of inputs. They are also the sort of tasks that can benefit from a consistent process, especially when teams want the same format every time.

Data cleanup is another example that fits the pattern. Many engineering and business workflows include low-value maintenance work, such as organizing records, checking for errors, or preparing information for later review. An automated assistant can take over part of that load, as long as the underlying instructions and guardrails are clear enough.

OpenAI has been pushing Codex toward a more practical role inside software teams, and automations extend that direction. Instead of treating the assistant as a place to ask questions, the company is presenting it as a system that can keep working on a timetable, which is the kind of behavior people usually expect from background jobs and workflow tools.

The company has also framed the feature around tasks that can be repeated rather than one-off creative work. That matters because scheduled automation only works well when the task definition is stable enough to run again and again without constant manual correction. In that sense, the announcement is less about flashy AI and more about making Codex behave like part of an operational routine.