OpenAI Rolls Out AI Agent ‘Operator’ In India

Amid the rise of agentic AI worldwide, GenAI giant OpenAI has now rolled out its AI agent “Operator” in multiple countries, including India, for its Pro users. 

In a post on X, OpenAI said, “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.”

Until now, Operator was only available to users in the US.  

Currently in a research state, Operator is an AI agent from OpenAI which can perform several tasks for the user. For instance, it can look at a webpage and interact with it by typing, clicking, and scrolling. 

The company claims it can handle repetitive tasks like filling out forms, ordering groceries, among others.

However, it is yet to be launched in the European Union (EU), Switzerland, Norway, Liechtenstein & Iceland – mostly due to the EU’s AI Act, which essentially regulates how AI companies operate, i.e data collection by internet scraping, facial recognition data collection, among others. 

The GenAI Model Powering Operator 

For Operator, OpenAI has created a new multimodal GenAI model called ‘Computer-Using Agent’ (CUA). The model combines GPT-4o’s vision capabilities and uses its advanced reasoning capabilities to interact with the screen. 

CUA processes raw pixel data to understand what’s happening on the screen and uses a virtual mouse and keyboard to complete actions. It can navigate multi-step tasks, handle errors, and adapt to unexpected changes. 

This enables CUA to act in a wide range of digital environments, performing tasks like filling out forms and navigating websites without needing any specialised APIs.

“If it encounters challenges or makes mistakes, Operator can leverage its reasoning capabilities to self-correct. When it gets stuck and needs assistance, it simply hands control back to the user, ensuring a smooth and collaborative experience,” OpenAI said in a blog post. 

To do a task, the model first adds screenshots from the computer screen for context, after which it reasons via a chain of thoughts process. It also takes into account the past and current screenshots. 

Notably, the agent is trained to decline sensitive tasks like banking transactions or those requiring high-stakes decisions, like making a decision on a job application.

Further, on sensitive sites like email or fintech platforms, the agent requires supervision of its actions. 

However, it only runs on the operating systems designed for such multimodal AI agents. 

The launch comes at a time when the usage of agentic AI is on the rise across the world. Deloitte predicts that 25% of companies operating in the GenAI field will launch agentic AI pilots or proofs of concepts in 2025, and this number would surge to 50% by 2027.

As a result, Indian IT companies are also bullish on agentic AI. In a recent post-earnings call, Infosys chief Salil Parekh said that the company is building over 100 GenAI agents for client applications in collaboration with its AI partner ecosystem.

Meanwhile, startups in the space are also seeing a lot of interest from investors. For instance, Gallabox raised $3.5 Mn recently to build agentic AI solutions. Meanwhile, Atomicwork bagged $25 Mn to launch agentic AI models.

In December, Gupshup CPO Gaurav Kachhawa, told Inc42 that more than 50% of its clients are actively exploring ways to use agentic AI.