Another week, another OpenAI launch. This one, though, is worth a closer look. ChatGPT Work arrived on 9 July 2026, and calling it a chatbot undersells it. It is an agent: hand it a brief and it comes back with the finished article, whether that is a spreadsheet, a slide deck, a report or a working web app.
ChatGPT Work at a glance
- Launched 9 July 2026, powered by the new GPT-5.6 model family
- Built into ChatGPT, with Codex technology under the surface
- Produces finished documents, spreadsheets, presentations and web apps
- Pulls context from connected apps such as Slack, Google Drive and CRMs
- Runs multi-step projects for hours, with plans and approvals you control
- Available globally in the desktop app on every plan; paid plans on web and mobile, with models and limits varying by subscription
What Exactly Is ChatGPT Work?
ChatGPT Work is an autonomous agent inside ChatGPT that gathers context across a user’s connected apps and files, breaks a goal into smaller steps and works through complex projects independently for hours at a time. Unlike standard ChatGPT, which drafts text inside a conversation, it delivers finished output such as spreadsheets, slides, documents and shareable web apps.
It has Codex technology built in, the coding agent more than five million people already use every week, and runs on the GPT-5.6 model family released the same day: Sol for heavy reasoning, Terra for everyday tasks (which OpenAI pitches at similar quality to its previous model for half the cost) and Luna where speed matters most. Before it starts, you see a plan you can review and adjust, and you decide what it can access, when it checks in and which actions need your approval. OpenAI’s launch announcement has the full rundown.
ChatGPT Work vs Standard ChatGPT
The simplest way to understand the difference is that one talks and the other delivers.
| Standard ChatGPT | ChatGPT Work |
|---|---|
| Answers questions and drafts text | Completes whole projects |
| Works within a single conversation | Runs long, multi-step workflows |
| Relies on what you paste in | Pulls context from connected apps and files |
| You prompt each step | It plans the steps and asks for approval |
| Output stays in the chat | Output is a finished document, sheet, deck or web app |
What Changed on Launch Day
On web and mobile, access began with Pro, Enterprise and Edu plans, with Plus and Business following in the days after launch. The bigger surprise sits on desktop. The Codex app has merged into a new ChatGPT desktop app for Mac and Windows that brings Chat, Work and Codex together on every plan, including Free, and the old app has been renamed ChatGPT Classic.
Two quieter changes came with it. Sites went into public beta, which lets you publish a piece of work as a shareable web app or live dashboard with its own URL. OpenAI also said it will start retiring Atlas, the standalone web browser it built around ChatGPT, with no date given yet, and move what it learned there into ChatGPT proper.
ChatGPT Work vs Claude Cowork
The timing here was no accident. Anthropic’s Claude Cowork has been doing this kind of agent work on desktop since January, and it expanded to web and mobile on 7 July, two days before OpenAI’s launch. Anthropic’s own usage data explains why both firms are chasing the same ground: more than ninety per cent of Cowork sessions are ordinary knowledge work rather than coding, with business operations and content creation the largest categories.
The practical differences right now come down to access. Cowork’s web and mobile beta starts with Anthropic’s Max plan and works down from there, while ChatGPT Work arrived across web, mobile and desktop in one go, with the desktop app covering every plan, including Free. For most businesses the sensible move is the same either way: pilot one tool on one workflow you know well, and measure it against the hours that workflow costs you today.
Why This Matters for Marketing and Business Teams
This launch fits a pattern we track closely through our AI search visibility work: the tools are now built to finish the job, not describe it. The question we ask clients about being found in AI-powered search now applies internally: whether your content, data and processes are structured well enough for an AI system to use them properly. That is the part the tech coverage keeps missing. An agent can only build a decent pipeline report or campaign analysis if your files, CRM records and brand guidelines are in order. Point it at a mess and it will produce confident rubbish. Consistent naming, clean CRM records and a single source of truth for brand assets have quietly become AI infrastructure.
The launch examples lean heavily on marketing workflows. Zapier’s head of enterprise marketing used ChatGPT Work to trace inbound leads across the company’s CRM and email and surface where follow-ups broke down, while Virgin Atlantic’s digital team used it to benchmark its customer experience against rival airlines, cutting an analysis cycle from weeks to hours. Shopify runs it as a daily working layer fed by Slack and live projects. All of these results are self-reported through OpenAI’s early access programme, so treat them as direction rather than benchmark.
There is a cost angle to plan for as well. Usage comes out of your plan’s allowance rather than being unlimited, and it follows the same structure as Codex: how much a task consumes depends on its size, its complexity and the model it runs on. A team that points Sol at everything in week one can empty a month’s allowance before anyone has measured whether the output was worth the spend.
Before introducing ChatGPT Work or a similar agent tool into daily operations, it is worth checking a few things first.
- Confirm whether your organisation has an approved policy for using agent tools on business data, rather than leaving the decision to individual staff
- Review exactly which connected apps and files an agent tool would be able to access before switching it on, particularly open Slack workspaces and shared drives
- Treat generated documents, decks and reports as drafts that need review, not finished output ready to send
- Keep a record of what was requested, what data informed it and who signed off before anything is sent externally
Frequently Asked Questions
What Does ChatGPT Work Actually Do?
ChatGPT Work takes on longer business tasks and hands back completed documents, spreadsheets, presentations and web apps, where a standard chat would stop at advice. It is OpenAI's attempt at an assistant that sees the job through.
When Did ChatGPT Work Launch?
ChatGPT Work launched on 9 July 2026, alongside the GPT-5.6 model family.
Who Has Access to ChatGPT Work?
On web and mobile, it is available on paid plans, starting with Pro, Enterprise and Edu, with Plus and Business added in the days that followed. In the updated ChatGPT desktop app for Mac and Windows, it is available on every plan, including Free.
Is ChatGPT Work Free?
It is included on the free tier within the ChatGPT desktop app, although models, features and usage limits vary by subscription. Web and mobile access requires a paid plan.
How Much Does ChatGPT Work Cost?
There is no separate fee. Usage comes out of a plan's included allowance and follows the same structure as Codex, so consumption depends on a task's size, complexity and the model used. Enterprise and Edu admins can set spend controls to manage it.
Is ChatGPT Work Available in the UK?
Yes. The updated desktop app launched globally, including the UK, and web and mobile access follows the standard plan rollout rather than a regional one.
Can ChatGPT Work Access Company Files?
Yes, through plugins that connect apps such as Slack, Microsoft Teams, Google Drive, SharePoint, email, calendars and CRMs. Users decide what it can access, and Enterprise and Edu admins can restrict connections and sensitive actions centrally.
Is ChatGPT Work Secure?
It is built on the security and workspace management foundation of ChatGPT Enterprise. Users approve sensitive actions, admins control access, connected tools and permitted actions centrally, and an auto-review layer checks important actions involving connected tools before they happen.
Can ChatGPT Work Build Websites?
Yes. The Sites feature, in public beta, turns a piece of work into an interactive site or web app shared through a URL, which suits dashboards, project trackers, prototypes and interactive reports.
Is ChatGPT Work Useful for Marketing Teams?
The earliest published use cases are marketing ones: lead triage, competitor benchmarking and turning customer research into campaign briefs and assets. Its value depends heavily on how well organised the underlying marketing data is.
Our View
Our reading is a simple one. The firms pulling ahead treat AI as infrastructure to plan around; the ones falling behind bolt it on afterwards. That applies as much to how your business appears in AI-powered search as it does to the tools your team uses internally, and in both cases the groundwork is the same: well-structured content, reliable data and processes an AI system can actually follow.
If you would like a second opinion on what any of this means for your own marketing, talk to our team.