Anthropic overtakes OpenAI, but CIOs stay put

Anthropic has overtaken OpenAI in U.S. enterprise AI adoption, capturing 34.4% of the market compared to 32.3%, according to the Ramp 2026 AI Index in May. Among first-time adopters, the Claude maker won roughly 70% of head-to-head matchups against competitors. The shift marks a milestone in the platform wars, yet for many CIOs, the leaderboard is not the main event. While vendor marketing emphasizes a single winner, the reality inside corporate IT departments is often more pragmatic and fragmented.

Security and governance come first

Before performance or pricing enters the conversation, CIOs say AI platforms must pass security and governance tests. At Cornerstone Research, Phil Leslie, the chief technology and innovation officer, emphasized that client data must never be used to train models and that interactions should not be exposed to human review. The firm’s data stays on U.S. infrastructure. “The differences among the leading frontier models are real but narrow, and they keep moving,” Leslie said. “The more useful question is not ‘which model is best this quarter’ but ‘which setup lets us switch as the frontier shifts.'” This focus on switchability suggests that even the market leader might not be the final choice for every enterprise.

Eric Pace of Cox Business echoed this sentiment, noting that security is non-negotiable given the critical infrastructure the company manages. The bar is just as high at the Lowenstein Sandler law firm. “Security and confidentiality aren’t one factor among several,” said Maureen Naughton, the firm’s chief information and innovation officer. “They are the threshold test.” For these organizations, the constraints define the feasible set, and everything else is a choice within it.

Building a portfolio of tools

With security requirements met, the decision process shifts to how to integrate multiple models rather than picking a single vendor. Cornerstone Research built a model-agnostic stack on purpose. “The frontier is moving too fast to wire our architecture to any single vendor; the lock-in risk is real,” Leslie said. Rather than betting on one winner, several CIOs describe running a portfolio of AI platforms matched to different use cases.

Lowenstein Sandler takes a similar approach, viewing different models as occupying distinct lanes rather than competing for one seat. This strategy is becoming the norm. “The continuous model leapfrogging has helped companies accept that the rate of change is only going to accelerate,” said Jeremy Bruck of West Monroe. Companies are no longer focused on the “smartest model” but instead on modular platforms that reduce switching costs as new solutions emerge. The evaluation process is also evolving. “The evaluation is moving from ‘Which AI vendor should we use?’ to ‘Which model should do this task, at what cost?'” said Ramp’s lead economist, Ara Kharazian. This shift pushes companies toward multi-vendor setups, routing, and open source models.

Developer preference plays a role in model selection, but it is filtered through a governance framework. Cox Business’ Pace describes this as “freedom within a framework.” The company provides a governed set of model options, and within that, teams have flexibility to choose what works best for their use case. “The key is that all of this happens inside a broader governance framework that we view as an enabler, not a constraint,” Pace said. Leadership’s job is not to overrule technical judgment but to ensure the whole evaluation actually happens.

Managing costs and agentic capabilities

As adoption scales, CIOs are developing new disciplines for managing consumption without stifling value. “AI is the fastest-growing spend category we’ve ever observed,” Kharazian said. The average business is spending 13x more on tokens than it was in January 2025. To manage this, West Monroe’s Bruck suggests tying spend to outcomes. “You can’t manage what you can’t attribute,” he said, “so disciplined enterprises track which team, workflow, and increasingly which person is driving spend, tied to a unit of business value.” Cox Business takes a value-first approach to cost management. “When AI helps someone work through years of backlog in weeks, the conversation shifts from controlling cost to asking what more we can enable,” Pace said.

At Lowenstein Sandler, Naughton noted that the bar has moved from novelty to durability. The criteria have matured toward integration depth, security posture, governance fit, and increasingly readiness for more agentic capabilities—tools that take actions rather than just generate text. Leslie of Cornerstone Research pointed out that once a system can search the web, call external services, or write and run its own code, the evaluation changes entirely. “What guardrails exist? What can we stop it from doing inside our environment?” he asked. The old test was about what the model produces, but the new test is also about what the model does. That is a genuinely harder problem. Model selection still matters, but for CIOs, so does building architecture that can adapt when the leaderboard flips again—because it will.