OpenAI Launches ChatGPT Agent That Can Book Flights & Fill Spreadsheets: Here's What We Know
OpenAI’s latest ChatGPT agent can now perform complex digital tasks like booking travel or organizing data, moving closer to hands-on AI-powered productivity.
OpenAI has introduced a new ChatGPT agent designed to independently carry out tasks on a user’s device, like booking flights, managing files, or filling in spreadsheets. The move is seen as a step toward more “autonomous agents,” AI systems that can take goals from users and complete multi-step actions across apps, all without continuous human input. The announcement was made during the company’s spring demo event on Tuesday, and the tools will be rolled out gradually in the coming months.
The updates will first go out to a small group of testers before becoming more widely available to GPT-4 users later this year.
Smarter, More Hands-On AI
Unlike older versions of ChatGPT, which mostly responded to queries or generated content, the new agent can operate software in real time. In OpenAI’s demo, the agent was shown browsing a website to book a flight, adding and sorting spreadsheet data, and pulling together expense reports, all without human clicks.
The agent uses what OpenAI calls a “software interpreter,” meaning it watches and acts on-screen the way a person would, instead of relying only on APIs. According to the company, this approach allows the agent to carry out more generalised tasks in a familiar interface, without needing apps to adjust how they function.
OpenAI CTO Mira Murati explained during the demo that this is part of making ChatGPT “much more useful in everyday work and personal tasks,” by letting it handle actual tools, not just provide instructions.
Gradual Rollout With Safety In Focus
While the prospect of AI that operates apps sounds powerful, it also raises concerns about safety, privacy, and control. OpenAI says it is releasing this feature slowly and with close oversight. Only a small group of testers will initially have access, and the system will come with guardrails to ensure it doesn’t misuse permissions or act unpredictably.
The goal is to refine the feature with feedback and safety checks before it becomes more widely accessible to GPT-4 users later in 2025.
As AI models evolve into agents that take initiative and handle entire workflows, the line between chatbot and digital assistant is starting to blur. For OpenAI, this launch is about showing that the future of AI is not just smart, it’s hands-on.
Top Headlines
