OpenAI Adds Team Tasks to ChatGPT: Cloud-Based Routines for Enterprise

OpenAI has released a video about team tasks in ChatGPT. These are team-wide tasks that run on a schedule in the cloud and are available to all members of a workspace. The feature is available only on Enterprise plans. As an example, the company used a legal department with a daily cycle: reviewing incoming requests, checking deadlines, and editing the shared instructions before the next run.

What OpenAI showed

The scenario is simple and familiar to any legal team. Every day, requests land in the department: approve a contract, check a wording, reply to an inquiry. A ChatGPT task picks up this queue, goes through the new requests, and checks them against deadlines. The result goes to a shared space where everyone can see it.

Then comes the most important part. A colleague opens the output of the run and assesses how it went. If the model set the wrong priorities or missed a type of request, the colleague edits the instruction. The next run uses the updated text. Over time, the task turns into a working procedure that the team refines as it goes.

OpenAI also stresses that tasks run in the cloud. For the process, this means it does not depend on an open browser tab, a laptop that is switched on, or a specific person being present. If the person who set it up is on vacation, everyone else can review the results and keep the work moving.

How this differs from personal tasks

Tasks in ChatGPT are not new. As far as is known, OpenAI launched Tasks in beta in January 2025, and at the time it was a personal tool: set a reminder, compile a summary, perform an action on a schedule. The new version takes the same idea to the team level. A task becomes a shared object with a run history and an instruction that anyone on the team can edit.

For companies, this addresses an old automation problem: knowledge of how a process works usually lives in the head of a single enthusiast. When that person leaves or moves to another project, the routine breaks and nobody understands why. A shared instruction kept in one place reduces that dependence on any one person.

What is known about the feature

• Format: team tasks, recurring tasks with shared management in ChatGPT.
• Availability: Enterprise plans.
• Execution: in the cloud, not tied to the user's device.
• Demo scenario: daily request intake in a legal department.
• Collaboration: viewing run results, editing shared instructions.
• Not disclosed in the material: pricing, run limits, integrations with external systems, details of logging and access permissions.

Questions the video does not answer

The video is marketing, not technical. It contains no accuracy metrics, no data on the time the legal department saves, and no description of errors. In a field like law, that matters. Incoming requests often contain confidential information, and a missed deadline costs money.

Before rolling this out, corporate customers should find out a few things. Who can edit the instruction, and is the edit history kept? What happens if a run ends with an error? Which data sources can the task access? How does it fit with information retention policies?

How to approach a pilot

It makes sense to start with a process that has a clear input, a clear deadline, and a cheap cost of error. Daily sorting of incoming tickets fits; preparing a final opinion for a client does not.

Write the instruction as you would a procedure for a new employee: what counts as urgent, which fields to check, and in what format to deliver the result. Assign a task owner who reviews every run in the first weeks and records the edits. It also helps to keep a short changelog of the instruction, so you can tell which edit improved the result and which made it worse.

And agree on a metric in advance: the share of requests handled without manual rework, or the number of missed deadlines per month. Without one, it will be hard to say after a quarter whether the automation works or just creates a feeling of order.