Ramp just published its May 2026 AI Index, and for the first time, Anthropic passed OpenAI in business adoption. The numbers: 34.4% of businesses on Ramp now pay for Anthropic, versus 32.3% for OpenAI. In the last year, Anthropic quadrupled its business customer base. OpenAI grew by 0.3%. But the headline is not who is winning. The headline is what this flip tells you about the companies that are actually using AI versus the ones that are not.

The Flip Is a Signal of Maturity, Not Loyalty

A year ago, OpenAI was the default. If your company was going to try AI, you signed up for ChatGPT. That was the starting line.

What the Ramp data shows is that the companies who started are now making their second and third decisions. They tried OpenAI. They evaluated alternatives. They switched to Anthropic because it fit their workflows better, or because Claude handled their specific use cases more reliably, or because the API economics worked out in their favor for the tasks they actually run.

That is not fandom. That is operational maturity. These are companies that have built enough internal practice with AI to know what they need and to act on it. They are optimizing, not experimenting.

The companies that have not started yet do not even have this problem. They are not choosing between providers. They are still choosing whether to try one at all.

The Gap No One Is Measuring

Every industry report tracks adoption rates. How many companies are “using AI.” The number is above 50% now, and it keeps climbing. That sounds like progress.

But adoption rate measures whether you started. It does not measure how far you have gone. The Ramp data reveals a different kind of gap: the distance between companies that started and iterated, and companies that started and stopped.

Overall AI adoption on Ramp rose 0.2 percentage points in April. That is almost flat. The market is not surging. What is happening is that the companies already inside are getting more sophisticated. They are switching providers. They are increasing spend. They are moving from “we have an AI tool” to “we have an AI strategy.”

The companies on the outside are standing still. Not because they said no. Because they said “not yet” twelve months ago and never revisited the decision.

Quadrupling Is Not Normal Growth

Anthropic went from roughly 8% to 34% of businesses in a year. That is not a marketing win. That is thousands of companies actively choosing to move their AI workflows to a different provider. They had to evaluate, migrate, retrain habits, update integrations. That takes effort.

The companies doing this are not the ones with the biggest budgets. According to the Ramp data, Anthropic’s growth is particularly strong among small and mid-sized businesses. These are teams with limited resources making deliberate decisions about where AI fits in their operations.

Meanwhile, OpenAI’s consumer business continues to grow. 900 million weekly active users on ChatGPT. $2.6 billion in monthly revenue. The consumer side is enormous. But the business adoption number barely moved. The enterprise push, the $4 billion deployment company, the consulting acquisitions: those are responses to the same problem the Ramp data surfaces. Getting people to sign up is not the hard part. Getting them to build real workflows is.

What This Means for You

If your company has not started with AI, the gap is no longer “they are ahead and we are behind.” The gap is “they are iterating on their second or third AI decision and we have not made our first.”

That is a different kind of distance. The first kind you can close with effort. The second kind compounds. Every month a company spends optimizing its AI workflows, it learns something. It builds muscle memory. It discovers what works for its specific operations. That knowledge does not transfer. You cannot buy it later. You have to earn it by doing the work.

The leaderboard flipped because the companies already moving were ready to make a better decision. The companies that were not moving did not even notice it happened.

The question is not whether your company should adopt AI. That question expired a year ago. The question is how many decisions behind you already are.