Thibault Sottiaux, who manages the token limits for OpenAI’s Codex tool, is now leading the charge for ChatGPT Work, a new platform designed for white-collar employees to use AI agents.
The product aims to extend the same growth experience users felt with coding tools to the broader workforce. Sottiaux told TechCrunch that the goal is to make agent technology safe and accessible for everyone, not just developers. The service is included in the Plus subscription plan, which costs $20 per month.
Reporting lines
Sottiaux leads the core product portfolio. This includes the API, agent infrastructure, enterprise solutions, all ChatGPT variants including the classic version, and the full Codex suite. He reports directly to Greg Brockman.
The economic case
Sottiaux argues that providing high utility drives willingness to pay. He stated that if a user receives significant value, they will naturally pay for that utility. The $20 monthly fee reflects this value proposition.
Diffusion strategy
OpenAI views the current moment as the right time to move technology from a technical audience to the general public. While early versions of Codex served a forgiving group of developers, the technology has reached a level of maturity suitable for a much wider population.
The focus is on autonomous task execution. Sottiaux explained that ChatGPT can now handle complex tasks entirely on its own in a safe and delightful manner. The mission remains to bring everyone along as the technology evolves.
Interface design
The product philosophy prioritises a minimal surface area. The aim is to let the model express itself without unnecessary buttons or interfaces. The goal is natural engagement where the user does not have to learn how to use the application.
ChatGPT Voice has seen significant growth, allowing users to speak naturally to the model. The progression is towards conversations that adapt to humans rather than requiring humans to adapt to the application.
Adoption rates
Sottiaux noted that the world appears ready for this level of automation. The platform launched with a focus on simplicity without compromise. OpenAI recently announced that it has reached 20 million users.
Product development
Development follows a discovery model. As the team pushes the frontier of model capabilities, they identify what the system can do best and build products around those strengths. This process involves capturing feedback and continuously improving the system based on real use.
GPT 5.6 serves as an example. The model can now process large document volumes, generate quality slides and reports, and conduct deep research. These are tasks professionals perform daily. The team leans into these specific capabilities to capture feedback and iterate.
Cost and efficiency
OpenAI is working daily to improve efficiency. A major price reduction of 80% was announced with the Luna update. This is described as a permanent price correction, meaning frontier capabilities will become cheaper over time.
The objective is to include more utility within the same dollar amount. Users should be able to achieve the same results in six months time with less spending.
Security and access
Users concerned about granting access to emails and iMessages can look for models that are safe and aligned. Sottiaux highlighted that a large part of OpenAI’s investment goes into the safety stack. The company publishes honest benchmarks to demonstrate that its models are world-class at handling these security topics.
What it means
For workers, the shift is from using AI as a simple assistant to deploying it as an autonomous agent for complex tasks. The $20 monthly fee buys access to features like deep research and document processing, while the 80% price cut on the Luna model ensures these capabilities remain affordable. The emphasis on voice interaction and minimal interfaces suggests a move toward tools that integrate into daily workflows without requiring technical training.
