OpenAI: AI Widens the Gap Between Leading Companies

The Rise of AI in Businesses: A Transformation Underway

Companies are adopting artificial intelligence (AI) at unprecedented levels, expanding not only its use but also the nature of the tasks assigned to these technologies. AI in business is no longer just a response tool; it is becoming an active participant in task execution. However, not all companies are progressing at the same pace. The most advanced companies, representing the top 10% in terms of monthly AI usage, now generate 8.3 times more output tokens per active user than typical companies. This figure reflects the intensity of usage and the growing gap between leaders and others, a gap that is widening with the adoption of capabilities that connect agents to the business context, tools, and repetitive work processes.

Two recent studies, "Enterprise Signals" and "How Organizations Use AI: Evidence from ChatGPT," explore this phenomenon. The first provides a practical perspective on agentic AI within OpenAI's enterprise clients, while the second examines the growth of AI adoption within companies, roles, and levels of seniority. Together, these studies highlight a practical agenda for businesses: connecting agents to the context and tools necessary to accomplish valuable work; establishing clear permissions, reviews, and governance; and helping employees transform effective individual workflows into shared working methods.

The Shift Towards Execution: AI Takes the Lead

AI in business is evolving from merely answering questions to executing complete tasks. Assistants help employees think through their work, while agents assist them in carrying it out. Tools like ChatGPT Work and Codex are capable of using tools, creating files, and producing work for review. For example, instead of asking the AI how to prepare a presentation, an employee can ask an agent to gather relevant information from various sources and draft the presentation itself.

This shift is evident in business usage. In June, Codex generated 64% of the combined output tokens of Codex and ChatGPT among enterprise clients. Agentic workflows generally produce more output because they accomplish longer, multi-step tasks, so this figure reflects both the frequency of Codex usage and the amount of output generated by these tasks.

A Growing Divide Between Companies

Every month, companies are ranked based on the number of output tokens per active user. Leading companies, those in the top 10%, generate 8.3 times more output tokens per active user than typical companies, which fall between the 45th and 55th percentiles. This gap is visible across various sectors and company sizes, indicating that intensive AI usage is not limited to tech companies. "Enterprise Signals" explores how this gap varies by sector and function.

Among the U.S. public companies studied in "How Organizations Use AI: Evidence from ChatGPT," AI adopters had stronger financial metrics compared to non-adopters. They held more assets, employed more workers, and had higher levels of R&D investment. These reports suggest that access to AI alone is not sufficient to expand its usage. Complementary investments in continuous employee learning, shared workflows, data infrastructure, and governance can support broader and deeper adoption.

Advanced Capabilities: An Asset for Leading Companies

For AI agents to be effective, they must have access to the right context and tools. Plugins bundle capabilities that help agents complete specific workflows. They can combine skills providing reusable instructions with applications connected to the company's data, tools, and actions. For example, a sales Plugin can combine a team's playbook with access to its CRM, allowing an agent to use current customer information and past proposals to prepare a personalized response for review.

Leading companies have a clear advantage in terms of advanced capabilities. Among weekly active users, 21% of leading companies use Plugins and 19% use skills, compared to only 9% and 3% in typical companies. However, adoption in leading companies represents only a fraction of what is possible. At OpenAI, 95% of employees use Plugins weekly, highlighting the potential for deeper utilization of these capabilities. Our research on how agents transform work provides a more detailed insight into how OpenAI employees use these advanced capabilities.

AI Expands into Knowledge Work

Software engineering has been an early center of agentic adoption, but the use of Codex is rapidly growing in knowledge work functions. Since February, the number of weekly active Codex users in enterprises has increased 108 times in the legal sector, 41 times in sales, 41 times in recruitment, and 26 times in marketing, compared to 5 times in engineering. At Virgin Atlantic, this shift is visible across the company. Engineering teams use Codex to refactor legacy code in 30 minutes instead of two weeks. Their product teams use ChatGPT Work to conduct weeks of competitive research in just a few hours, shaping the company's digital strategy for the next five years. "Enterprise Signals" explores how these capabilities are spreading across sectors and functions, drawing on a sample of over 10 million messages.

Young Employees: The Vanguard of AI Adoption

Contrary to some surveys, administrative data from millions of conversations reveals that AI usage is higher among early-career employees than among executives. Six months after adoption, these young employees sent 13 more messages per week than executives. For leaders, this finding underscores a practical opportunity to identify employees with the strongest AI habits and make their workflows visible, thus helping effective practices spread across all levels.

Bridging the Gap: A Challenge for Leaders

Although companies have access to the same AI models, leading companies implement them more quickly and deeply within their organizations. The opportunity for leaders is to bridge the gap between leading companies by extending agentic workflows beyond engineering and transforming successful individual use cases into repeatable practices at the organizational scale. By connecting agents to the context and tools of the business, with clear permissions, governance, and human review, companies can shift from assistance to execution.