Artificial intelligence is shifting from text-based assistance to action-driven AI agents capable of executing digital tasks. OpenAI has expanded the availability of Operator, an AI-powered automation tool within ChatGPT Pro, allowing users to complete structured online tasks such as filling out forms, retrieving information, and handling web-based workflows.
The rollout brings Operator to ChatGPT Pro subscribers in Australia, Brazil, Canada, India, Japan, Singapore, South Korea, the U.K., and more countries.
Operator is now rolling out to Pro users in Australia, Brazil, Canada, India, Japan, Singapore, South Korea, the UK, and most places ChatGPT is available.
— OpenAI (@OpenAI) February 21, 2025
Still working on making Operator available in the EU, Switzerland, Norway, Liechtenstein & Iceland—we’ll keep you updated!
With Operator , OpenAI positions itself in direct competition with companies investing in AI-powered task automation. Google is preparing to launch a similar product with Project Mariner, an AI agent built with Gemini 2.0 which “can understand and reason across everything on your browser screen, including pixels and web elements like text, code, images and forms” for more complex automation.
Anthropic’s Claude 3.5 Sonnet already extended its AI capabilities into full desktop control, enabling automation of entire operating system workflows across Windows and macOS.
Operator AI: How It Works and Where It Stands
Unlike conventional AI chatbots that rely on user prompts for each action, Operator enables semi-autonomous digital task execution. It interacts with web-based systems, automating structured processes like appointment scheduling, data retrieval, and web form completion. However, OpenAI has strictly limited its level of independence.
Operator requires explicit user approval before executing any action, and a manual takeover mode allows intervention if needed. By restricting Operator to web-based automation, OpenAI is taking a cautious approach compared to Anthropic’s Claude 3.5 Sonnet, which enables full desktop automation, including UI interactions, keyboard input, and system navigation.
The Competitive AI Agent Landscape
AI-powered task execution is quickly becoming a battleground for tech companies. Google, OpenAI, and Anthropic are competing with different approaches to AI-driven automation. Microsoft is also entering this arena and just unveiled its foundational AI model Magma which also enables AI agents managing software and even robotic systems via multimodal input.
Google’s project Mariner and Operator also use multimodal AI, which allows them to process text, images, and code in parallel, making them more effective for complex workflows, while being focused on web-based tasks.
The transition from chatbots to AI agents capable of executing tasks has been developing for years. OpenAI, Google, and Anthropic have gradually advanced their models, refining automation capabilities and improving AI-driven decision-making.
In December 2023, OpenAI internally tested early AI agent prototypes. By January 2024, the introduction of ChatGPT’s memory feature allowed AI models to retain past interactions, setting the foundation for context-aware automation. A few months later, Google, Microsoft, and Meta publicly introduced their own AI-driven automation strategies.
Anthropic’s Claude 3.5 Sonnet’s Computer Use feature became the first AI tool capable of controlling entire Windows and macOS environments, paving the way for full desktop automation.
OpenAI’s Operator is the latest sign that AI-driven automation is moving beyond experimental phases and into practical applications. As AI agents evolve, the industry will need to determine how much autonomy AI should have, whether automation should remain strictly user-supervised, or if AI-driven task execution will gradually expand into more independent decision-making models.
