OpenAI is planning to turn ChatGPT into a “super app”—an all-purpose AI interface that brings OpenAI’s key products together—as it heads toward an upcoming IPO. At the center of that push is Thibault Sottiaux, the former head of Codex, who has recently been promoted to oversee OpenAI’s core product platform for consumer, business, and developer customers.
At the VivaTech technology conference in Paris, Sottiaux told Fortune that the company planned to evolve ChatGPT from a conversational tool into a more personal assistant that remembers what users care about and gets better the more it learns from them. The end goal, he said, is a single, highly capable agent that can handle both simple requests and more complex tasks—from answering quick questions to planning a complex trip and then booking all the travel arrangements to spinning up a custom app to help their child learn trigonometry—all in one place.
The company plans to merge ChatGPT and Codex into one unified agentic experience. Over the past few months, it has been reorganizing its product leadership to achieve that. Product strategy is now under OpenAI president Greg Brockman with Sottiaux leading core product and platform, and Nick Turley—previously head of ChatGPT—moving to lead enterprise industries. The consolidation also came with the shutdown of various side projects, including Sora, its AI video product, which was discontinued in April. The company also pulled back from expensive infrastructure commitments, halting plans for Stargate data center projects in the U.K. and Norway.
Sottiaux, who was also instrumental in building OpenAI’s popular Codex developer tool, said the company planned to use Codex as the engine powering the new super app.
“Codex has evolved now to a place where it is very general, and it is not specific to code anymore. We’re working on using all of that and bringing that directly to ChatGPT, so that it can benefit a very broad user base,” he said.
The company has made various attempts at agentic products before. But tools like Operator, the company’s earlier attempt at an agentic product that could access the internet and take actions on the web on a user’s behalf, like browsing sites, filling forms, and completing tasks autonomously, largely failed to take off with users.
“You can think about Operator and understand it was early,” Sottiaux said. “It didn’t quite work, because the models were not” ready. The difference now, he argued, is that the models can handle a much larger set of tools, parse genuinely ambiguous instructions, and operate with more permissions because of advances in ensuring the models accurately follow user intentions. Operator often had to operate in a constrained “sandbox” because it would make mistakes if given wider latitude to conduct transactions on a user’s behalf.
Super apps are ubiquitous in China and much of Southeast Asia—platforms like WeChat and Alipay that bundle messaging, payments, food delivery, ride-hailing, and commerce into a single destination, so that users rarely need to leave the app to conduct daily life. Western tech companies have largely failed to replicate these kinds of apps, in part because users in the U.S. and Europe are accustomed to switching between specialized apps.
OpenAI’s vision, according to Sottiaux, is less about replicating that model and more about building what he called a “personal AGI”: a unified interface that grows more capable and more personalized to a user over time. “It’s about bringing more capabilities in one very simple unified interface that enables you to communicate, steer, and supervise your personal AGI,” he said.
OpenAI is also trying to sharpen its case against Anthropic in the enterprise market—a fight that has shifted slightly in Anthropic’s favor over the past year.
Sottiaux says the company does not see a hard divide between consumer and enterprise use cases when it comes to the new product, but there is a question of whether enterprise customers would prefer tailored, vertical solutions built for specific industries and workflows over an all-in-one app. Large enterprise buyers have typically been suspicious of horizontal platforms that promise to do everything, favoring integrations that slot neatly into existing systems.
“We think that fundamentally the technology doesn’t look that different,” Sottiaux said, adding that the same system can be useful in both personal and work settings because it is about “connecting the right context, connecting the right tools, and then taking actions in a way that minimizes risk in order to deliver value.”
