Late last week, OpenAI made a surprise announcement unveiling its new, general purpose AI agent integrated into ChatGPT. According to the blog post announcing the tool, ChatGPT agent takes the best parts of its previously released agentic efforts (Operator and deep research) and combines them in a “unified agentic system.”
OpenAI’s Pro, Plus, and Team users can select “agent mode” from a dropdown box to activate the functionality.
“Operator and deep research each brought unique strengths: Operator could scroll, click, and type on the web, while deep research excelled at analyzing and summarizing information,” Open AI wrote in the announcement. “By integrating these complementary strengths in ChatGPT and introducing additional tools, we’ve unlocked entirely new capabilities within one model. It can now actively engage websites—clicking, filtering, and gathering more precise, efficient results. You can also naturally transition from a simple conversation to requesting actions directly within the same chat.”
The company acknowledged that the tool “is still in its early stages,” that it still makes mistakes and its outputs are not polished. Several outlets highlighted some of the mistakes the system is capable of (e.g., placing a baseball stadium in the middle of the Gulf of Mexico) and even the project’s research lead Isa Fulford admitted in the video accompanying the release that it took ChatGPT agent an unreasonably long time to make a simple cupcake order.
It does, however, “expect continued improvements to ChatGPT agent’s efficiency, depth, and versatility over time, including more seamless interactions as we continue to adjust the amount of oversight required from the user to make it more useful while ensuring it’s safe to use.”