ChatGPT Work Is Live: What OpenAI’s New Work Agent Actually Does

OpenAI shipped ChatGPT Work on July 9, 2026. Skip the launch-day headlines. Here are the actual example use cases pulled straight from OpenAI’s own product page, so you can see what it’s actually built to do.

ChatGPT Work pulls from your apps, files, tools, browser, and approved work context to produce documents, presentations, spreadsheets, websites, reports, and analyses.

OpenAI positions ChatGPT Work as a new agent that gathers context, builds a plan, produces a polished output, and keeps the project moving. Three things stand out in how OpenAI itself describes it.

Set a goal, schedule something recurring, or ask it to monitor for updates. It gathers context, builds a plan, and checks in with you when a decision or approval is needed.

Start, review, and direct tasks from web and mobile. Switch to desktop for deeper workflows that need local files, apps, browser access, or Codex.

You supply the context, pick the outputs, decide what actually gets used, sent, shared, or automated. It asks before acting and documents what it did with references.

One output type worth flagging: websites. ChatGPT Work can spin up a simple site or dashboard as part of a task. Fine for an internal tracker. A completely different job from a properly built customer-facing website, worth knowing before anyone assumes this replaces that.

These are the exact use cases OpenAI lists on the ChatGPT Work product page, not paraphrased marketing copy.

Weekly brief
“Turn my calendar, messages, and documents into a weekly brief covering priorities, risks, decisions, and follow-ups.”

Survey feedback
“Analyze this survey feedback using Google Sheets. Extract the key themes, representative quotes, and open questions, and update the team summary.”

Email replies
“Identify the emails I need to reply to. Summarize each one, then draft replies and leave them unsent for my review.”

Launch brief
“Turn the product doc, launch timeline, and Slack thread into a launch brief with owners, risks, decisions, and next steps.”

Attendee list
“Build an attendee list from these leaders, draft the invitations, create a tracker, and flag anything that needs manual review.”

Training follow-up
“Check this training spreadsheet to find employees under 100% completion, then draft direct Slack messages using the template in this document.”

Pipeline review
“Compare this quarter’s sales pipeline spreadsheet against last week’s forecast, then summarize the changes, risks, assumptions, and follow-up actions.”

Project tracker
“Build a project tracker from this planning document, identify likely owners based on the discussion, then draft a status update for the team channel.”

The pattern across all eight: feed it real, messy inputs (calendars, spreadsheets, Slack threads, email) and ask for one specific, finished output. None of them are vague “help me with marketing” prompts. That specificity is worth stealing no matter which agent you’re using, and it works just as well for marketing workflows as it does for the ops examples OpenAI lists.

ChatGPT Work is an agent that plugs into your existing apps and turns a stated goal into a finished document, spreadsheet, or tracker, asking permission as it goes. The prompts above are OpenAI’s own, and they’re a genuinely solid template for how to prompt any agent: specific inputs, a named output, a clear scope.