OpenAI is bringing its powerful Codex AI agent technology to non-technical workers through ChatGPT Work, a new platform that has already attracted 20 million users. The shift represents a major turning point for the company: what started as a tool for software engineers is now being packaged as a general-purpose assistant for white-collar professionals. This expansion reflects OpenAI's belief that the technology has matured enough to safely serve a much broader audience.

Why Is OpenAI Pushing Codex Beyond Coders?

Thibault Sottiaux, who leads all of OpenAI's core products including Codex and ChatGPT Work, explained the strategic reasoning behind this shift. "We wanted to bring the power of coding agents to everyone, and so this is an exercise in taking something that was made for technical people, and then packaging it in a way that is like safe and delightful to use," Sottiaux said in an interview with TechCrunch.

"We wanted to bring the power of coding agents to everyone, and so this is an exercise in taking something that was made for technical people, and then packaging it in a way that is like safe and delightful to use, but also you can use on the go on mobile, on web, and making it available to as broad of a population as possible," explained Thibault Sottiaux.

Thibault Sottiaux, Head of Product at OpenAI

The company launched ChatGPT Work as part of its Plus subscription plan at just $20 per month, making it accessible to a wide audience. OpenAI's economic motivation is clear: the more utility the platform provides to users, the more they will be willing to pay for it. This mirrors the company's broader strategy with ChatGPT, where users perceive the value as far exceeding the monthly cost.

What Can ChatGPT Work Actually Do for Everyday Workers?

Unlike earlier versions of Codex that required users to understand technical concepts, ChatGPT Work is designed to handle entire complex tasks autonomously. According to Sottiaux, the platform can now perform sophisticated work that previously required human effort, including processing large volumes of documents, generating quality slides and reports, and conducting deep research.

The key difference from earlier AI tools is the level of autonomy. Rather than requiring users to prompt the AI repeatedly or understand how to structure requests, ChatGPT Work can take on multi-step tasks independently. This represents a significant maturation of AI agent technology, moving from a tool that assists with individual tasks to one that can manage entire workflows.

How Does OpenAI Think About Releasing Powerful AI to the General Public?

OpenAI's approach to expanding Codex reflects a deliberate philosophy about product design and user interface. Rather than overwhelming users with buttons and options, the company is stripping away complexity to let the AI model express its capabilities naturally. Sottiaux described this as "getting out of the way" of the model so users can harness its full utility.

Sottiaux

The company has also expanded how people interact with AI beyond text. ChatGPT Voice, which launched alongside these new capabilities, has seen significant adoption by allowing users to have natural conversations with the AI, similar to how people talk to each other. This progression toward more natural interfaces is expected to continue as the technology evolves.

"The essence of what we're trying to do is building extremely capable models, and then figuring out the most simple and delightful way to bring them into your life so that you get tremendous utility from it," stated Sottiaux.

Thibault Sottiaux, Head of Product at OpenAI

Steps to Understand ChatGPT Work's Capabilities and Limitations

• Autonomous Task Handling: ChatGPT Work can independently manage complex workflows like document processing, slide generation, and research compilation without requiring step-by-step user guidance for each subtask.
• Natural Interaction Methods: Users can engage with the platform through text or voice, with the system designed to adapt to human communication patterns rather than requiring users to learn specific technical syntax or commands.
• Safety and Alignment Focus: OpenAI emphasizes that its models undergo rigorous safety testing and alignment work, with the company publishing honest benchmarks on safety topics to ensure the AI behaves responsibly when given access to sensitive information like email and messages.
• Iterative Improvement Process: The platform learns from real-world usage and community feedback, with OpenAI continuously refining capabilities based on what users actually need rather than what the company initially predicted.

What About Cost and Efficiency Concerns?

One question many potential users have is whether the $20 monthly subscription truly covers the cost of running such powerful AI. Sottiaux acknowledged this concern but pointed to OpenAI's ongoing work to improve efficiency. The company recently announced major price cuts with its Luna model, offering 80% discounts on frontier capabilities. These are permanent price reductions, not temporary promotions.

Sottiaux's message to both individual users and corporate finance officers is straightforward: costs will continue to decline. "Our goal is to, over time, include more utility in the same dollar amount," he explained. This means that six months from now, users should be able to accomplish the same work with less spending, even if they want to do more.

How Is the Broader AI Infrastructure Adapting to Support Agents?

Beyond OpenAI's own products, the entire ecosystem is shifting to support AI agents. A new startup called Keenable, backed by venture capital firm Accel, is building specialized web search infrastructure specifically designed for AI agents rather than humans. The company has indexed over 100 billion documents and is already being used by multiple AI labs and inference providers during both training and runtime.

Andrey Styskin, Keenable's co-founder and former head of search at Russian search giant Yandex, explained that AI agents need different infrastructure than traditional search engines. "If you do not fine-tune your index structures for a specific task, the cost of serving and scanning the whole internet is enormous because of the volume," Styskin noted.

Andrey Styskin, Keenable's co-founder and former head of search at Russian search giant Yandex

"If you do not fine-tune your index structures for a specific task, the cost of serving and scanning the whole internet is enormous because of the volume. That's why you need to innovate on how you can narrow the search space based on your query very fast. This is what we are bringing to the table," explained Andrey Styskin.

Andrey Styskin, Co-founder of Keenable

This infrastructure shift reflects a broader recognition that AI agents operate differently from human users. While humans can quickly scan search results and pick the most relevant one, AI agents can process much larger amounts of information if it's properly indexed and structured. Keenable is developing tools like Web Query Language to help AI systems answer questions by combining information from multiple web sources, even when no single source contains the complete answer.

The timing of these developments suggests that the AI industry believes the technology has reached a critical maturity point. OpenAI's expansion of Codex to general workers, combined with infrastructure investments like Keenable's specialized search index, indicates that AI agents are transitioning from experimental tools to mainstream productivity platforms. Whether this transition succeeds will depend on whether the technology can deliver on its promise of autonomous, reliable task completion while maintaining user trust and safety.