OpenAI's product chief: the world is ready for AI agents that do your work

OpenAI's head of product believes the moment has arrived to hand over serious work tasks to AI agents. In an interview, Thibault Sottiaux describes the strategy behind ChatGPT Work—the company's push to make autonomous AI tools accessible to white-collar workers at $20 per month—and how he thinks about the skeptics.

From specialists to everyone#

Sottiaux oversees OpenAI's core products, from APIs and agent infrastructure to ChatGPT in all its forms and the Codex software engineering tool. His core mission with ChatGPT Work is taking capabilities originally built for technical users and repackaging them for a mass audience. The shift reflects confidence that the underlying technology has matured enough for deployment beyond expert hands.

The framing is deliberate: rather than positioning this as a specialized tool, Sottiaux emphasizes universal adoption. "Make this technology available to the broadest population possible," he says, describing the philosophy behind the $20-a-month price point and mobile-first experience.

Agents that work autonomously#

What distinguishes ChatGPT Work from earlier versions of the company's chatbot is the degree of autonomy. The platform can now handle entire complex workflows—processing documents, generating presentations and reports, performing research—with minimal human intervention. This represents a step beyond writing assistance or advice-giving; it's about outsourcing substantial professional tasks.

Sottiaux sees this as the right time for such a shift. The company has hit 20 million users on the Work product, which he frames as evidence that "the world seems to be ready" for this level of agent capability. The success suggests people are willing to trust AI with meaningful work responsibilities when the interface is simple and the output is reliable.

The interface question#

A key design challenge has been deciding when to let the AI lead versus when to expose controls to the user. Sottiaux describes the philosophy as "getting out of the way"—building minimal UI and letting the model do what it's trained to do. Recent launches like ChatGPT Voice exemplify this: users speak naturally, as they would to a human, rather than learning application mechanics.

This contrasts with competitors. Ethan Mollick, a Wharton professor studying AI tools, observed that Anthropic's Claude Cowork puts A/B tests and choices in front of users, while OpenAI leans into what it calls "magic"—autonomous execution that users may or may not understand in detail.