From chatbot to colleague: what ChatGPT Work really is
ChatGPT Work is an autonomous ChatGPT Work agent that uses GPT-5.6 capabilities and Codex to run multi-step task automation across connected apps, files, and web services, turning a single written request into hours of coordinated work that looks far closer to an AI employee than a typical chatbot. OpenAI launched ChatGPT Work as an agent that can research, build documents, manage spreadsheets and create web applications for hours at a time without constant human supervision. It can complete multi-step projects using information from connected applications, online services and local files, dividing big goals into smaller tasks and producing spreadsheets, presentations, documents and web apps. Crucially, this is not about faster answers to prompts; it is about delegating entire workflows—monitoring connected apps, running scheduled jobs and producing finished deliverables that slot into existing enterprise AI workflows.
How GPT-5.6 turns goals into multi-step task automation
The strategic bet behind ChatGPT Work is that reasoning plus coding is the real unlock for enterprise AI workflows. GPT-5.6 powers the agent, giving it the ability to reason through multi-step tasks, follow templates and reference files, and keep context across long chains of work. Sol, the flagship GPT-5.6 model, costs USD 5 (approx. RM23) per million input tokens and USD 30 (approx. RM138) per million output tokens, Terra costs USD 2.50 (approx. RM11.50) input and USD 15 (approx. RM69) output, while Luna costs USD 1 (approx. RM4.60) input and USD 6 (approx. RM27.60) output. OpenAI claims Sol delivers state-of-the-art results across coding, cybersecurity and science while using fewer tokens than competing frontier models, and that GPT-5.6 can create tasteful, functional interfaces from high-level direction alone. Codex is built in, allowing the agent to write code, use tools and coordinate actions as projects develop.
What autonomous enterprise AI workflows look like in practice
The meaningful change is not a smarter chat; it is a different way of working. Users can give the ChatGPT Work agent a goal—analysing a monthly budget, preparing for a sales meeting or turning research into a campaign brief—and it divides the assignment into smaller tasks, gathers information and produces spreadsheets, presentations, documents and web apps. Users can assign an entire workflow through one request: a project might start with customer research, move into a campaign brief and continue into localized materials, all while retaining what it learned in earlier stages. The agent stays active on longer projects so humans can review progress, answer questions or approve actions mid-stream. It also supports Scheduled Tasks that review new messages, refresh meeting documents, check sites or dashboards for changes and update presentations when new feedback arrives, turning recurring to‑do lists into background processes.
The new integration layer: apps, files, desktops and Sites
ChatGPT Work matters because it sits on top of the digital stack instead of replacing it. The agent connects to Slack, Microsoft Teams, Google Drive, SharePoint, Gmail, calendars, CRM platforms and project trackers, pulling from information already scattered across an organisation. A user can point it to a specific system or let it choose the best connected source for the goal. The updated desktop app adds access to local files, applications and a built-in browser that can gather information from websites and work with Google Workspace and Microsoft 365 files. Computer Use lets the agent click, type and move files across supported apps and browser tools, effectively automating the tedious glue work humans usually perform when switching between services. Sites, now in public beta, lets ChatGPT Work turn project outputs into interactive websites or lightweight web apps that can be tested in ChatGPT and shared via URL, then kept up to date as underlying data changes.
An open battle with Claude Cowork—and what happens next
OpenAI is not subtle about the competitive target: ChatGPT Work is a direct answer to Claude Cowork, another agent that plans and runs multi-step tasks. Both products aim at people who are not coders but want the muscle of coding tools, which often outperform standard chatbots yet demand specialized skills. Both companies are preparing for possible public offerings and are fighting for enterprise customers, who spend more than consumers. The timing is shaped by politics as much as product; GPT-5.6 and ChatGPT Work launched after a delay tied to national security concerns, following earlier reporting that the release had been held up by the Trump administration. ChatGPT Work is available now on web and mobile for Pro, Enterprise and Edu plans, with Plus and Business to follow in the coming days, while Free and Go users are excluded from the agent. The GPT-5.6 models are rolling out over 24 hours and GPT-5.4 will be retired on July 23, though GPT-5.5 models remain. In effect, OpenAI is collapsing its previous Operator, deep research and Workspace Agents efforts into a single, long-running work agent that can research, write, code and orchestrate tasks for hours at a time. That fusion—code generation plus task orchestration—sets the new bar for enterprise AI workflows, and it is where the agent race will be decided.
