ChatGPT Work Turns AI From Chatbot Into Autonomous Worker

From Chatbot to Co-worker: What ChatGPT Work Actually Is

ChatGPT Work is an autonomous AI agent from OpenAI that runs multi-step tasks for hours—researching, building documents and web apps, managing spreadsheets, and coordinating software workflows—without constant human supervision, while still letting users monitor, redirect, and approve its actions as it works. OpenAI launched ChatGPT Work during a livestream event, framing it less as a friendly chatbot and more as a digital employee that can take a vague goal and grind through the tedious work in the background. That framing matters: the company is not selling a smarter conversation partner; it is selling time. By automating the glue work of modern knowledge jobs—data gathering, formatting, cross-app coordination—ChatGPT Work challenges our assumption that AI should sit in a chat window waiting for prompts. Instead, it occupies the same space as human assistants and junior analysts: autonomous, accountable, and expected to deliver finished output.

Agentic AI Task Automation: How It Works Across Apps and Devices

The bold claim behind the ChatGPT Work agent is that it can take on jobs that usually stretch across days of human effort, compressing them into hours by running unattended. Designed for multi-step tasks, it breaks large goals into smaller actions and shuttles data through connected tools like Slack, Microsoft Teams, Google Drive, and Salesforce. One enterprise customer example is telling: Virgin Atlantic had the agent compare its passenger experience against competitors, build a dataset, and highlight where the airline led or lagged—a process the company says cut weeks of analysis down to hours. This is AI task automation in a serious sense: not a macro recording a few clicks, but an autonomous AI agent orchestrating research, synthesis, and deliverable creation at scale. If this works reliably, many "busywork" tasks stop being human tasks at all and become oversight functions—reviewing, editing, and deciding what to do with what the agent produced.

Native on iPhone and Mac: The Agent Lives Where You Work

OpenAI wisely refused to keep ChatGPT Work trapped in a browser tab. On Apple hardware and desktop platforms, you can start a multi-hour project from an iPhone, watch drafts evolve on iOS, and then pick up the same workflow on a Mac or Windows PC without friction. That level of native integration matters because serious automation has to live inside daily tools, not outside them. The updated desktop app merges the developer-focused Codex environment directly into ChatGPT, combining Chat, Work, and Codex in one interface for macOS and Windows. This unification quietly reframes app building automation: generating code, editing repositories, and turning results into web applications all happen inside a single agentic workspace rather than a patchwork of IDEs and browser-based tools. The result is an AI worker that feels less like an add-on and more like part of the operating system of your job.

GPT-5.6 Under the Hood: Power, Pricing, and Platform Consolidation

ChatGPT Work runs on GPT-5.6, OpenAI’s latest frontier model family, released alongside the agent after a limited preview. The lineup comes in three tiers: Sol, Terra, and Luna, with Sol positioned as the flagship. Sol is priced at USD 5 (approx. RM23) per million input tokens and USD 30 (approx. RM138) per million output tokens through the API, Terra at USD 2.50 (approx. RM12) input and USD 15 (approx. RM69) output, and Luna at USD 1 (approx. RM5) input and USD 6 (approx. RM28) output. OpenAI claims Sol delivers state-of-the-art performance in coding, cybersecurity, and science while needing fewer tokens than competing frontier models. GPT‑5.6 also improves the quality of artifacts—presentations, documents, spreadsheets—and works better with existing templates, making exported outputs more usable in real workflows. The company is rolling these models out over 24 hours and plans to retire GPT-5.4 on July 23, while keeping GPT-5.5 models available. This is not a casual update; it is a platform reset around long-running agents.

Computer Use, Sites, and the Future of Autonomous AI Agents

The real shift is not only that ChatGPT Work can think for hours; it can act. A desktop feature called Computer Use lets the assistant execute background keyboard and mouse commands on a user’s computer to automate routine operations. Combined with a built-in browser that replaces the experimental Atlas product, the agent can gather web information and manipulate cloud files directly. Meanwhile, a Sites beta feature lets users turn work into interactive websites or lightweight web apps without leaving ChatGPT, then publish them by URL. To make this scalable, Pro and Enterprise users get an ultra acceleration mode that coordinates multiple agents in parallel workstreams to finish complex tasks faster. Access is gated—ChatGPT Work launches on web and mobile for Pro, Enterprise, and Education tiers today, with Plus and Business users following, while Free and Go users remain excluded from the agent itself. That choice signals where OpenAI expects the value: in organizations ready to treat autonomous AI agents as part of their workforce, not as a novelty chatbot.