OpenAI Launches Operator: The Future Where AI Can Perform Online Tasks for You Is Here

  • Operator can handle online shopping and manage restaurant reservations.

  • It’s currently only available for ChatGPT Pro users in the U.S.

OpenAI's Operator
No comments Twitter Flipboard E-mail
javier-marquez

Javier Márquez

Writer

I've been in media for over a decade, but I've been much longer marveling at the possibilities that technology brings us. I believe we live in a world where the digital revolution is changing everything, and I find no better palce that Xataka to write about it. LinkedIn

OpenAI introduced ChatGPT in 2022. In 2023, the company released GPT-4, which featured an innovative Her-style voice mode. The following year, 2024, saw the debut of the o1 reasoning model. On Thursday, OpenAI announced Operator, which aims to become one of the top products of the year.

CEO Sam Altman and his team demonstrated live how Operator, which places AI agents at the forefront, works. The tool can currently perform some actions within a web browser, but the goal is for it to undertake more complex tasks over time.

OpenAI’s Big Bet on Agents

OpenAI’s goal with Operator is to assist users with various tasks, including online shopping, booking flights, and making restaurant reservations. For example, if you wanted to book a table, you could say, “Book me a table for two at Beretta for 7 p.m. tonight.” Operator will then browse the web to fulfill your request.

If any issues arise, such as unavailability at the chosen time or the need for sensitive information, Operator will prompt you for input. It may ask you to select a different time or location, or it might request personal details and payment methods to complete the reservation.

Similar to ChatGPT, Operator also offers customized instructions. This feature allows you to set very specific preferences for tasks you perform regularly, such as monthly grocery shopping.

OpenAI’s Operator interface OpenAI’s Operator in action.

An intriguing feature of Operator is that it comes with its own built-in browser, eliminating the need for users to install additional software or browser extensions. At its core, Operator utilizes a new model called the Computer-Using Agent (CUA), which merges the visual capabilities of GPT-4o with advanced reasoning through reinforcement learning.

This allows the system to “see” what is displayed on your browser screen and interact with its elements, such as buttons, dialog boxes, text fields, and navigation bars. Additionally, the model can “self-correct” when necessary, which helps it navigate challenges and theoretically provides a smoother user experience.

Operator has the ability to “self-correct.”

Operator is still in the research phase, which means this is just an initial release. OpenAI has made a preliminary version available to the public at operator.chatgpt.com. However, access is limited. It’s currently available only to ChatGPT Pro users ($200 per month) in the U.S.

Additionally, it’s important to note that the Operator is still in full development. As such, the company cautions that the system may make mistakes and may struggle with complex interfaces. OpenAI also announced that it’s collaborating with companies such as Uber, DoorDash, OpenTable, and Instacart to enhance Operator’s performance.

Image | OpenAI

Related | OpenAI Is Developing a Ph.D.-Level AI Model. The Product Is So Impressive That It’s Already Landed a Meeting With the U.S. Government

Home o Index