OpenAI: enterprise pivot

OpenAI said this week that enterprise customers now bring in more than 40% of its revenue, and it expects enterprise to reach parity with consumer revenue by the end of 2026. Two months earlier, it launched OpenAI Frontier, a platform built to help companies deploy and manage artificial intelligence agents inside their own systems. (openai.com, openai.com) That is a different business than selling chat boxes to employees one seat at a time. Frontier is pitched as an “intelligence layer” that connects company data, permissions, and software so agents can act across finance, support, operations, and other internal workflows. (cnbc.com, openai.com) An artificial intelligence agent is software that does a job in steps instead of answering one prompt and stopping. OpenAI’s own developer tools now emphasize stateful conversations, file search, web search, computer use, and function calling so the model can keep context and use outside systems. (developers.openai.com, openai.github.io) “Agentic workflows” means several of those agents split up a task the way a team would. One agent can gather documents, another can check policy rules, and a third can write the final output, with the company controlling permissions and review points around them. (openai.com, decrypt.co) That changes how companies buy the product. A seat license is a budget line for employee software, but a workflow system touches security reviews, data access, procurement, compliance, and the teams that own the underlying software being connected. (decrypt.co, cnbc.com) It also changes who gets to influence the deal. When the sale is a chatbot, the buyer is often an information technology team or a business unit leader, but when the sale is an agent platform, systems integrators, consultants, and internal architecture teams suddenly matter because they decide what can be wired into payroll, customer records, or supply-chain tools. (decrypt.co, openai.com) OpenAI has been laying the groundwork for that shift for a while. In June 2025, it said it had more than 3 million paying business users across ChatGPT Enterprise, ChatGPT Team, and ChatGPT Edu, giving it a large installed base to upsell from simple workplace chat into deeper automation. (cnbc.com) The company’s own numbers show why it wants developers and corporate buyers thinking in platforms, not prompts. OpenAI says its application programming interfaces now process more than 15 billion tokens per minute, and it says GPT‑5.4 is driving record engagement across agentic workflows. (openai.com) The customer list it highlighted also fits that story. OpenAI named Goldman Sachs, Philips, and State Farm as new customers, and said existing customers including DoorDash and Thermo Fisher are expanding, which suggests the pitch is landing in regulated and operations-heavy industries where automation budgets are larger than chat budgets. (openai.com) So the headline is not just that enterprise is a bigger slice of OpenAI’s revenue now. It is that OpenAI is trying to move from being an app employees open in a browser to being a layer companies build into the machinery of how work gets approved, executed, and audited. (openai.com, openai.com)