Just a few minutes after unveiling its new GPT-5.6 family — Sol, Terra, and Luna — OpenAI is back with another announcement. This time, it’s introducing ChatGPT Work, a new AI agent designed to do more than answer questions. Instead of helping with one task at a time, it can take on entire projects that span multiple apps, documents, and services.
If you’ve ever spent an afternoon jumping between different apps just to finish a single assignment, ChatGPT Work is trying to eliminate that back-and-forth. The idea is to describe the end goal and figure out the steps in between. This represents a fundamental shift in how we interact with AI systems: from a reactive chatbot that waits for each command to a proactive agent that can plan, execute, and adapt across a complex workflow.
Give it a goal, then get out of the way
Powered by GPT-5.6, ChatGPT Work can connect to the apps and files you choose, gather the information it needs, and produce completed work instead of just suggestions. That could mean drafting a report, building a spreadsheet, creating presentation slides, or even putting together a simple web app without you having to micromanage every prompt. The underlying technology leverages OpenAI's Codex system, which has been trained on vast repositories of code and natural language instructions, allowing the agent to understand and execute multi-step tasks across different software environments.
OpenAI says the experience works best with real-world jobs people already spend hours doing. Imagine asking it to review last month’s business expenses or gather everything you need before an important client meeting. Rather than starting from scratch every few minutes, you can watch the agent work, answer questions as needed, redirect it midway, or approve actions before it proceeds. The company is also pitching ChatGPT Work as capable of handling long, connected workflows. For example, it could analyze customer feedback, turn those findings into a marketing strategy, generate the supporting assets, and then adapt everything for different audiences without losing track of what happened in earlier steps.
This capability is not entirely new in the AI landscape — Anthropic's Claude Code recently introduced an in-app browser that can browse websites, click links, and interact with web apps. However, ChatGPT Work goes a step further by integrating with multiple services simultaneously and maintaining context across disparate tasks. The agent is built on the latest GPT-5.6 architecture, which offers improved reasoning, longer context windows, and better handling of complex instructions. The Sol, Terra, and Luna variants represent different tiers of capability and cost, with ChatGPT Work likely utilizing the most powerful Luna model for demanding workflows.
Your AI doesn’t clock out when you do
Perhaps the most interesting feature is that it doesn’t necessarily stop when you close ChatGPT. Using Scheduled Tasks, Work can continue running in the background, checking connected services like Slack or Microsoft Teams for new information, updating documents or presentations automatically, and notifying teammates when something important changes. In theory, you could assign work before leaving the office and come back to a project that’s already moved forward. That’s a notable shift from the AI assistants most people are familiar with today.
This background processing capability opens up new possibilities for workflow automation. For instance, a marketing team could set up a weekly task where ChatGPT Work analyzes website analytics, pulls in social media engagement data, drafts a report, and posts it to a shared drive — all without human intervention. The agent can also handle exception cases by sending a notification or pausing to ask for clarification. This reduces the cognitive load on employees who currently spend hours on repetitive data gathering and formatting tasks.
The concept of an AI agent that can run autonomously is not without risks. OpenAI has implemented approval gates for certain actions, such as sending emails or making financial transactions. Users can configure the level of autonomy: full automatic mode, semi-automatic with confirmation prompts, or manual step-by-step approval. This granular control addresses concerns about AI systems making costly mistakes or violating data privacy. Additionally, the agent operates within a sandboxed environment that restricts its access to only the apps and files explicitly authorized by the user.
Competitors are also moving in this direction. Microsoft has been integrating AI agents into its Copilot suite for Microsoft 365, allowing similar cross-app automation within the Office ecosystem. Google is working on Project Mariner, an experimental agent that can browse the web and complete tasks on behalf of users. However, OpenAI's approach with ChatGPT Work is more open-ended, supporting not just Microsoft and Google services but also third-party apps like Notion, Asana, Jira, and custom APIs through plugin integrations.
Under the hood, ChatGPT Work uses a combination of GPT-5.6's language understanding and a new orchestration layer that maps user goals to sequences of API calls and interface interactions. When the agent encounters an app that doesn't have an API, it can use computer vision and simulated clicks — similar to the technique used by Claude Code's in-app browser. This makes it compatible with legacy software and web applications that lack modern integration options.
Early beta testers have reported mixed results. While the agent excels at well-defined tasks like data extraction and report generation, it sometimes struggles with ambiguous instructions or scenarios that require subjective judgment. For example, asking it to "improve the design of a presentation" might lead to unexpected color choices that don't align with brand guidelines. OpenAI is actively collecting feedback and plans to release regular updates to improve the agent's reliability.
The pricing structure for ChatGPT Work is tiered: Pro subscribers ($200 per month) get unlimited usage, while Enterprise and Edu customers have dedicated quotas based on their organizational needs. Free and Plus users can test a limited version with reduced capabilities and a cap of 10 automated tasks per month. This model mirrors the strategy behind GitHub Copilot’s enterprise tiers, where the most advanced AI features are reserved for high-paying customers.
ChatGPT Work is rolling out starting today on the web and mobile apps for Pro, Enterprise, and Edu subscribers, with Plus and Business users expected to gain access over the coming days. OpenAI is also releasing an updated ChatGPT desktop app for both Windows and macOS, bringing Chat, Work, and Codex together in one place. While the desktop app is available across all ChatGPT plans, including the Free tier, access to the new Work agent itself will depend on your subscription.
The launch comes at a critical time for OpenAI. The company is facing increasing competition from Anthropic, Google, and open-source alternatives like Llama. By introducing an agent that can handle long-running, multi-app workflows, OpenAI is positioning itself as a provider of AI that can actually do real work — not just answer questions. The success of ChatGPT Work will depend on how well it navigates the balance between autonomy and control, and whether it can deliver on the promise of saving hours of manual labor each day. As more businesses explore AI agents for productivity, the ability to seamlessly integrate with existing tools and processes will be the key differentiator.
Source:Digital Trends News