OpenAI leans into enterprise

OpenAI says business customers now bring in more than 40% of its revenue, and the company says that share could match consumer revenue by the end of 2026. It is a sharp change for a company that first became famous selling one chatbot to millions of people at a time. (openai.com) The new pitch is not “here is a model, go build on it.” The new pitch is software that can plan steps, call tools, keep state, and finish multi-step office work more like a junior employee than a search box. (openai.com) OpenAI has been laying the plumbing for that shift for more than a year. Its Responses application programming interface is now the recommended starting point for new projects, and OpenAI describes it as a unified interface for building agent-like applications with built-in tools. (openai.com) This is different from the first wave of corporate artificial intelligence spending, which often meant buying chat access for staff. ChatGPT Enterprise launched in August 2023 with security, privacy, long context windows, and administrative controls aimed at large companies rather than consumers. (openai.com) Now OpenAI is saying the center of gravity has moved again. In its April 2026 enterprise update, it said its application programming interfaces process more than 15 billion tokens per minute and named Goldman Sachs, Philips, and State Farm among customers using its business products. (openai.com) The split inside the market is becoming easier to see. Newer cloud-native companies can plug agents into modern software stacks faster because their data already lives in connected systems and their workflows are easier to expose through application programming interfaces. (redhat.com) Older enterprises have a harder job because the work starts before the model does. They often need to clean data, map permissions, add governance rules, and connect old systems that were built for fixed transactions rather than autonomous software making decisions across departments. (kpmg.com) That is why implementation firms are suddenly central to the story. OpenAI said in February 2026 that it had formed multi-year partnerships with McKinsey, Boston Consulting Group, Accenture, and Capgemini to help deploy and govern its Frontier platform for enterprise agents. (theailibrary.co) It also helps explain why companies that sell outsourcing and integration work are talking so much about “agentic” systems. Mint reported in March 2026 that newer firms are adopting faster, while legacy enterprises are slower because integration, governance, and data cleanup are the real bottlenecks. (livemint.com) OpenAI is betting that the next big software budget will not go to raw model access alone. It will go to systems that can open the ticket, check the policy, pull the file, ask for approval, and finish the task inside the company’s own workflow. (openai.com) If that works, the winners will not just be the companies with the smartest model. They will be the companies with the cleanest data, the safest permissions, and the fewest old systems blocking the agent from doing the job. (servicenow.com)