ChatGPT Work can spend hours on a project after a single request, marking OpenAI’s clearest shift yet from chat responses to long-running workflow execution inside ChatGPT.
OpenAI has begun rolling out the new agent for work that is more complex than a typical prompt, according to PYMNTS. The agent can act across a user’s apps and files, keep working when the user is away from a computer or phone, and produce finished materials such as sheets, slides, docs, web apps and other outputs.
OpenAI rolls out ChatGPT Work agent for complex tasks across apps and files
OpenAI said in a Thursday (July 9) blog post that ChatGPT Work is powered by GPT-5.6, a new frontier model the company also rolled out Thursday. That model is the engine behind the agent’s push beyond short answers and into multi-step work.
The company’s examples are pointed at office tasks that already have clear workflows: budget variance analysis, marketing briefs and sales preparation. OpenAI is not pitching this as a smarter autocomplete. It is pitching a system that can take a whole assignment and move it forward.
“The best way to learn how to use ChatGPT Work is to give it a task you already know well: analyze a month-end budget variance, turn source materials into a marketing campaign brief, or prepare for a sales meeting,” OpenAI said in the post. “You can follow its progress, answer questions, change direction and approve important actions.”
The approval line matters. A chatbot that drafts a paragraph is one thing. An agent that can take action across apps and files is a different risk profile, because the work may touch documents, calendars, repositories or other business systems depending on what the user connects.
Rollout is phased:
| Platform | Initial availability | Later availability |
|---|---|---|
| Web and mobile app | Pro, Enterprise and Edu plans on Thursday | Plus and Business plans over the next few days |
| ChatGPT desktop app | Available on every plan | Not specified |
OpenAI CEO Sam Altman also tied the launch to the broader GPT-5.6 release. He said Thursday that one of the new models, GPT-5.6 Sol, is 54% more token efficient on agentic coding jobs and is “as good or better” than competing models on the market.
“Every enterprise now is thinking about spend and the value they’re getting in exchange for AI, and this is what we really want to do,” Altman told CNBC Thursday.
ChatGPT Work pushes OpenAI deeper into enterprise productivity software
XOOMAR analysis: ChatGPT Work is OpenAI’s move to make ChatGPT useful after the brainstorming phase ends. The product is aimed at the work between apps: gathering context, sorting source material, creating a deliverable and asking for approval when needed.
That middle layer is where office work gets expensive. Not because each step is hard, but because the steps are scattered across files, messages, meetings and documents. OpenAI’s pitch is that ChatGPT Work can hold the thread across that mess for longer than a normal chat session.
The company had already signaled this direction. PYMNTS reported in June that OpenAI was overhauling ChatGPT into a “super app” by bundling AI agents, coding tools and third-party services into a single platform. Thibault Sottiaux, who oversees all of OpenAI’s core product and platform, told the Financial Times that OpenAI wants to build “your own personal agent that is capable of helping you across everything in your life, be it personally or at work.”
For readers tracking OpenAI’s broader push beyond simple chat, XOOMAR has also covered how ChatGPT is affecting commerce behavior in ChatGPT Shoppers Crush Search Traffic for Retailers and the company’s legal pressure points in 400 Apple Defectors Ignite OpenAI Lawsuit Over ChatGPT.
The enterprise angle is obvious, but it is not automatic. Businesses will weigh speed against permissions, data access, reliability and audit trails. A long-running agent is only useful if teams can see what it did, stop it when needed and trust the output enough to send it into a real workflow.
Long-running AI agents raise questions about control, accuracy and data security
The risk level rises when an AI system stops only suggesting and starts acting. ChatGPT Work can take action across apps and files, according to OpenAI, which means users and IT teams will want sharper answers about scope.
The practical questions are simple:
- Access: Which apps and files can the agent read?
- Actions: What can it change without approval?
- Review: Can users inspect the steps before accepting the result?
- Recovery: What happens when the agent gets stuck or makes a bad assumption?
- Administration: How much control do organizations get over permissions and usage?
OpenAI’s quote says users can follow progress, answer questions, change direction and approve important actions. That is the right framing, but the details will decide whether companies treat ChatGPT Work as a productivity tool or as another system that needs heavy supervision.
Accuracy is the other pressure point. If ChatGPT Work creates a financial spreadsheet, sales prep document or web app, errors may be harder to spot than a bad sentence in a draft. A small mistake can move from a file into a meeting, then into a decision.
XOOMAR analysis: The agent will be judged less by its most polished demos and more by its worst ordinary tasks. Can it finish the boring work cleanly? Can it explain why it made a choice? Can it ask for help before it breaks the assignment? Those are the questions that separate a useful agent from a risky assistant.
OpenAI's next test is whether ChatGPT agents can deliver finished work at scale
The rollout phase now becomes the real test. Users will not grade ChatGPT Work on model benchmarks alone. They will judge it by whether it produces usable spreadsheets, decks, documents and apps without creating cleanup work.
The immediate watch items are concrete: who gets access first, how the tool behaves across web, mobile and desktop, what app integrations are supported, how file permissions work, what controls Enterprise and Edu customers receive, and whether usage limits shape how often people can run long jobs.
Pricing also remains a practical question. The source material says which plans receive access and when, but it does not spell out any separate pricing, caps or usage terms for ChatGPT Work. That leaves businesses to watch for whether the agent becomes a standard feature inside paid plans or a reason to move up tiers.
OpenAI’s bigger signal is clear. The AI race is shifting from answering questions to completing work. If ChatGPT Work can reliably carry projects for hours, OpenAI will have moved ChatGPT closer to the center of daily office execution. If users find they must constantly correct it, the agent will become another inbox instead of a way out of one.
Why It Matters
- OpenAI is moving ChatGPT from a conversational tool toward an agent that can complete substantial work assignments.
- The ability to act across apps and files could make ChatGPT more useful for business workflows but also raises approval and security concerns.
- GPT-5.6 positions the new agent as a major step beyond short-form AI assistance into autonomous productivity tasks.
Originally published on XOOMAR. For more news and analysis, visit XOOMAR.
Top comments (0)