OpenAI announced Operator today, its first attempt at an AI-powered agent that can automate complex tasks and perform various actions on websites to save you time.
This includes making restaurant reservations, shopping online, and booking travel accommodations. Specific sensitive actions may require user approval. “On particularly sensitive websites, such as email, Operator requires active user supervision, ensuring users can directly catch and address any potential mistakes the model might make,” OpenAI explains. This is the reason why Operator currently doesn’t support sending emails or deleting calendar events, but OpenAI is working on that. There are automations in task categories like Delivery, Dinning, Shopping and Travel. Explanations of specific actions being utilized are displayed on the screen while Operator is performing automation.
Instead of using developer APIs to plug into web apps, Operator’s Computer-Using Agent (CUA) model has been trained to interact directly with website frontends using its own dedicated web browser. OpenAI claims Operator respects the terms of service agreements of its launch partners DoorDash, eBay, Instacart, Priceline, StubHub, and Uber.
The ChatGPT maker doesn’t expect the CUA to perform 100% reliably all the time.OpenAI’s support documentacknowledges as much, saying, “Operator cannot reliably handle many complex or specialized tasks.” Some examples include “creating detailed slideshows, managing intricate calendar systems, or interacting with highly customized or non-standard web interfaces.” Operator has other downsides, including task-specific rate limits and an overall usage limit that resets daily. Moreover, it can fail in some tasks, like solving a CAPTCHA challenge, and has difficulty navigating complex web interfaces.
OpenAI earlier implementedsimple automation capabilities in ChatGPT, like setting reminders, but Operator is its first attempt at an AI agent. Rival Google unveiled its own AI agent in November 2024,Project Mariner, as an experimental Chrome extension that can fill out web forms on your behalf, click buttons, move the mouse pointer, and more.
AI agents are considered the next logical step in the AI revolution. These things promise to use the web on your behalf based on your prompt, freeing you from directly interacting with websites. However, the utility of AI agents is currently questionable, at best, as they’re in experimental stages and won’t be used widely until reliability improves.
