From Chatbot to Co-Worker: What ChatGPT Work Really Is
ChatGPT Work is an AI-powered workplace automation tool that acts as a multi-step task agent, using GPT-5.6 and Codex to plan, execute, and coordinate complex workflows across connected apps, online services, and local files.
This is not another chat interface; it is OpenAI’s bid to turn ChatGPT into an enterprise AI agent that can own outcomes, not just conversations. Users give it goals like analysing a monthly budget, preparing for a sales meeting or turning research into a campaign brief, and it breaks the work into tasks, pulls data from tools, and produces spreadsheets, presentations, documents and even web apps. In other words, ChatGPT Work automation shifts the system from “answer my question” to “finish this project.” That is the headline change: AI moves from sidekick to semi-autonomous colleague, and that will pressure every existing workplace automation tool to get smarter fast.
How Multi-Step Task Agents Change Day-to-Day Work
The real disruption is not that ChatGPT Work can write another slide deck; it is that it can own an entire multi-step pipeline. A user can assign an entire workflow through one request: start with customer research, continue into a campaign brief, then spin out localized materials while reusing everything learned earlier in the process. Unlike a traditional chatbot, it is designed to manage complex, multi-step tasks, break large projects into smaller actions and keep working even when users are away.
This directly targets the “glue work” knowledge workers spend hours on: hunting for files, copying data between tools, reformatting updates. Scheduled Tasks let it review new messages, refresh meeting documents, monitor dashboards and update presentations when new feedback arrives. One quotable fact stands out: “More than five million people use Codex each week, including more than one million who use it for work outside software development.” Those users now get a far broader automation canvas, pushing routine coordination work out of human calendars and into AI queues.
GPT-5.6, App Connections, and the New Enterprise Stack
OpenAI is clear about its ambition: GPT-5.6 is the engine, ChatGPT Work is the agent, and together they aim straight at the heart of enterprise automation. ChatGPT Work is powered by the latest GPT-5.6 model and Codex technology, using that combination to reason through longer workflows and create materials that follow templates and reference files. Alongside it, OpenAI released the GPT-5.6 family—Sol, Terra and Luna—and claims GPT-5.6 Sol slightly outperforms Anthropic’s Claude Mythos 5 in a coding benchmark.
Connections turn this model into a practical workplace automation tool. ChatGPT Work plugs into Slack, Microsoft Teams, Google Drive, SharePoint, Gmail, calendars, CRM platforms and project trackers, drawing on information already spread across the organisation. The updated desktop app adds local files, applications and a built-in browser that can work with Google Workspace and Microsoft 365. With Computer Use, it can click, type and move files across supported apps, effectively acting as an RPA-lite agent without the brittle scripts. This stack blurs the line between classic bots, RPA, and AI copilots—exactly where enterprise AI agents were always headed.
Competing with Claude and Classic Automation Platforms
Context matters: ChatGPT Work arrives months after Claude Cowork triggered intense debate about AI’s role in the workplace and its impact on IT teams. OpenAI is not pretending this is a generic upgrade; it is positioning ChatGPT Work, powered by GPT-5.6, as a direct rival to Claude Cowork for workplace automation. That places it squarely in the agentic AI space, up against both AI-native coworkers and incumbent enterprise automation platforms.
This is where the strategy becomes obvious. One quotable statement from OpenAI’s disclosures is that “OpenAI has unveiled ChatGPT Work, a new AI agent built into ChatGPT that is designed to handle workplace tasks across apps, files and workflows.” Add Sites—now in public beta, turning project data into interactive web apps like dashboards, trackers and internal portals—and you get an end-to-end story: agents execute the work, and Sites keep the results live. The message to enterprises is blunt: why stitch together separate workflow tools, RPA bots and dashboard builders when one AI agent can coordinate them all?
Governance, Rollout, and the Hard Part: Using It Wisely
Granting an AI agent access to company documents, messages, customer records and cross-app actions raises a simple question: how much autonomy is too much? OpenAI’s answer is governance knobs. Users decide what the agent can access and when approval is required, while Enterprise and Edu administrators manage connected tools, browser use, network resources and sensitive actions. An auto-review system checks important actions involving connected tools and APIs, and a Compliance API gives visibility into ChatGPT Work conversations and activity.
Rollout is already under way: ChatGPT Work is available on desktop across all plans, while web and mobile access is in a phased launch to Pro, Enterprise and Edu users, with Plus and Business plans following shortly. The opportunity is clear—offload repetitive, multi-step work to AI agents—but so is the risk. Wider access amplifies the impact of bad instructions, excessive permissions or incomplete information. The conclusion is straightforward: ChatGPT Work automation is powerful enough to reshape how teams operate, but only organisations that pair it with disciplined governance and thoughtful task design will benefit instead of regret it.
