OpenAI Says Enterprise AI Has Moved Beyond Experimentation

OpenAI used an April 8 note from Chief Revenue Officer Denise Dresser to argue that enterprise AI has moved beyond pilots and into deployment. The company attached several concrete metrics to that claim: enterprise now represents more than 40% of revenue, Codex has reached 3 million weekly active users, OpenAI's APIs process more than 15 billion tokens per minute, and ChatGPT now has 900 million weekly users.

Those figures matter because OpenAI is trying to position itself as more than a model supplier. In the note, the company said enterprise customers are asking two practical questions: how to put the most capable AI to work across an entire business instead of isolated copilots, and how to make AI part of daily work for every team. OpenAI's answer is to present its next phase as an agent platform strategy rather than a collection of individual assistants.

From point tools to company-wide agents

OpenAI described OpenAI Frontier as the layer it wants customers to use for building, deploying, and managing agents across a whole organization. The company pointed to customers such as Oracle, State Farm, and Uber, and to implementation partners including McKinsey, BCG, Accenture, Capgemini, AWS, Databricks, and Snowflake. It also highlighted a Stateful Runtime Environment being built with AWS so agents can keep context, remember prior work, and operate across multiple internal systems and data sources.

That positioning is notable because many enterprise AI products over the last year have focused on chat interfaces, departmental assistants, or narrow automations. OpenAI is now explicitly arguing that large buyers want fewer disconnected AI products and more centralized governance, shared context, and cross-system execution. The company's language suggests it expects the next buying cycle to favor vendors that can combine models, orchestration, connectors, permissions, and end-user interfaces in one stack.

What the note signals

OpenAI also tied enterprise adoption to consumer familiarity. It said ChatGPT's 900 million weekly users lower deployment friction because many employees already understand the basic interface and workflow. Combined with rising Codex usage and stronger agentic workflows, the message is that the limiting factor is no longer interest in AI, but how quickly organizations can connect capable models to real business processes. This was not a single new product launch. It was a clear statement that OpenAI intends to compete as full-stack enterprise AI infrastructure, with agents at the center of that pitch.