News
OpenAI Launches ChatGPT Work, an Agent for Autonomous Office Tasks
OpenAI has introduced ChatGPT Work, an agent built on the GPT-5.6 model that independently plans and executes multi-hour office tasks for businesses, drawing on data from Slack, Gmail, Google Drive and CRM systems.
OpenAI has launched ChatGPT Work, a new feature that turns the chatbot into an agent capable of independently planning and carrying out complex, multi-hour office tasks. It's the company's latest move in the race to dominate the enterprise AI market, where it now competes with Anthropic's Claude Cowork as well as tools from Microsoft and Google.
Unlike a traditional chatbot that responds to single commands, ChatGPT Work is designed to understand a goal set by the user, independently draw up a plan of action and carry it out step by step without constant supervision. The tool can prepare a quarterly report, run a competitive analysis, organize documentation, build a spreadsheet, or even create a simple web application.
What the new agent can do
A key feature of ChatGPT Work is its integration with the tools companies use every day. The agent connects to Slack, Gmail, Google Drive, CRM systems, company files and internal knowledge bases to gather the data it needs on its own, then combines information from multiple sources into finished business documents.
The feature runs in a redesigned ChatGPT desktop app available for Windows and macOS, as well as in the web and mobile versions. OpenAI also introduced a safeguard called Auto-review, which in tests conducted by external red team groups blocked every attempt to extract protected, confidential information from the system.
The GPT-5.6 model behind it
ChatGPT Work runs on the GPT-5.6 family of models, which OpenAI describes as offering stronger multi-step reasoning, better adherence to templates and more effective use of source material. The company released three variants of the model: Sol, the most powerful; Terra, balanced for performance and cost; and Luna, optimized for speed, letting companies choose the model that fits a given task.
Every enterprise is now thinking about spend and the value they're getting back for AI - Sam Altman, OpenAI CEO
OpenAI is also highlighting the growing scale of enterprise adoption of its agentic tools. Codex, the coding agent built into ChatGPT, has already surpassed 5 million weekly users, with more than a million of them using it for tasks not directly related to programming, for example finance teams automating month-end reporting and forecasting.
The race for the workplace
The launch of ChatGPT Work comes just months after Anthropic introduced its own office agent, Claude Cowork, with similar capabilities for planning and executing multi-step tasks. Microsoft, Google and Salesforce are also joining the fight for enterprise customers, making office-work automation one of the most fiercely contested fronts in the competition among AI providers.
For companies in Poland, this adds another tool to weigh when choosing an AI provider for administrative, finance or legal departments, where lengthy, multi-step tasks such as preparing reports or analyses currently consume the most employee time. OpenAI also says it plans further integration of Codex into the desktop app and an expansion of the feature to include a Chrome browser extension in the coming months.
Sources: Bloomberg (bloomberg.com), Forbes (forbes.com), The Tech Portal (thetechportal.com)

