OpenAI Rolls Out ChatGPT Work Agent for Complex Tasks

OpenAI has started rolling out a new agent in ChatGPT designed for multi-step work, not just Q&A. The tool, called ChatGPT Work, can take actions across a user's apps and files, keep working when the user walks away, and produce finished materials like sheets, slides, docs, and web apps, the company said in a Thursday blog post.

The agent runs on GPT‑5.6, a new frontier model released the same day. OpenAI CEO Sam Altman told CNBC that one version, GPT‑5.6 Sol, is 54% more token efficient on agentic coding jobs and “as good or better” than competing models. “Every enterprise now is thinking about spend and the value they’re getting in exchange for AI, and this is what we really want to do,” Altman said.

The launch puts OpenAI’s agent directly against Anthropic’s Claude Code, Google’s Gemini agents, and Microsoft’s Copilot tools. The difference is scope: ChatGPT Work can manage entire workflows, not just single tasks. Thibault Sottiaux, who oversees OpenAI’s core product and platform, told the Financial Times the company wants to build “your own personal agent that is capable of helping you across everything in your life, be it personally or at work.”

Pricing and availability will determine adoption speed. ChatGPT Work is rolling out first to Pro, Enterprise, and Edu plans on web and mobile, then to Plus and Business plans over the next few days. Desktop app users get it on every plan. That wide distribution could pressure rivals to match the capability or lower prices, said enterprise software analysts who follow the space.

GPT‑5.6’s claimed efficiency gains matter for cost-conscious buyers. If Sol really is 54% more token efficient on coding, enterprises running heavy agent workflows could see lower compute bills when using OpenAI’s platform. That would strengthen the argument for standardizing on OpenAI rather than running multiple model providers, the analysts said.

For public‑company shareholders, the read‑through runs through Microsoft, OpenAI’s largest investor and exclusive cloud partner. Each new OpenAI tier that drives more API calls or Azure compute usage adds to Microsoft’s AI revenue line, while competing models from Google or Anthropic face harder sales conversations if OpenAI’s agent proves stickier with existing customers.

Altman’s comment about “spend and value” speaks to the current enterprise mood. Budgets are tightening after two years of rapid AI piloting. Agents that demonstrably replace human work – month‑end variance analysis, marketing campaign prep, sales meeting preparation – offer a clearer ROI case than general‑purpose chatbots. The test will be whether ChatGPT Work actually closes those workflows without requiring constant human correction.

OpenAI has not disclosed usage numbers or revenue from the new agent. The phased rollout lets it fix issues before full exposure. “You can follow its progress, answer questions, change direction and approve important actions,” the company said in its announcement. That suggests the agent is not fully autonomous yet – it still needs humans in the loop for key decisions.