Key facts
- OpenAI is developing AI agents designed to perform complex, multi-step tasks autonomously.
- The company has released ChatGPT Work, a product aimed at white-collar professionals, available for $20 per month.
- The goal is to enable AI agents to integrate into digital workflows for various professions, including finance and law.
- OpenAI's lead engineer, Andrew Ambrosino, is testing AI agents with extensive access to his digital life, including email and Slack.
- A study indicated low adoption rates for agentic coding tools among external subscribers compared to internal employees.
- OpenAI is focusing on making AI agents more general-purpose and intuitive for non-technical users.
OpenAI is developing AI agents designed to perform complex, multi-step tasks autonomously, aiming to expand their use beyond software engineers to professionals in fields like finance, law, and medicine. The company's ChatGPT Work product, launched last month and available for $20 per month, is central to this strategy.
Andrew Ambrosino, lead engineer for OpenAI’s desktop app, is testing these agents by granting them access to his inbox, Slack, phone, and applications like Notion and Figma. He acknowledges the potential risks of AI agents accessing private information but states he has not encountered issues. This approach is part of OpenAI's effort to make AI tools more accessible and valuable to a broader audience, moving beyond simple question-answering to enabling users to realize their ideas.
While AI has proven lucrative for software engineers through tools like Codex, its adoption in other departments has been slower. ChatGPT Work is intended to provide non-engineers with similar capabilities, allowing AI to handle entire projects autonomously. Thibault Sottiaux, who leads OpenAI's core product work, emphasized that this aligns with OpenAI's mission to bring everyone along.
Commercially, agents that operate for longer durations consume more tokens, increasing revenue for OpenAI. Reaching new professions is crucial for the industry to justify massive investments in AI training and computation. While major AI labs focus on software engineers, specialized competitors like Harvey for law and Clay for sales are pursuing customers with a model-agnostic approach.
Industry analysts note that if AI labs cannot secure key complementary assets for market scaling, value may accrue elsewhere. OpenAI is working to make its AI tools more general-purpose, moving away from code-specific interfaces to a more intuitive, prompt-based experience. An internal study revealed that while 98% of OpenAI employees used Codex, only 17% of organizational subscribers and less than 1% of individual subscribers were using the agentic coding tool.
To bridge this adoption gap, OpenAI is focusing on user experience, comparing its approach to skeuomorphism, which helped users transition to new digital tools. The company aims for a "magic box" interface, making complex AI functionalities as easy as simple prompting. While the exact user numbers for Work versus Codex were not disclosed, the joint app is used by 20 million people, a fraction of ChatGPT's over one billion online users.
OpenAI is currently pitching these AI agents for routine, data-intensive coordination tasks, such as generating weekly metrics reports or creating data visualizations from Slack conversations. Professionals are using these agents to assemble company information into investment memos and to spin up bespoke dashboards. The goal is to help users parse vast amounts of information and take action, with ChatGPT providing access to this data within existing systems.