Anthropic recorded the fastest enterprise AI growth among knowledge-worker tools over a four-year period ending June 2026, according to new Okta data. But the real headline isn’t just who’s growing fastest — it’s that 57% of organizations now use at least two AI platforms, shattering the illusion of a single-vendor future. For Windows administrators, this means managing a multi-platform AI stack has become an immediate operational challenge, even as Microsoft 365 remains the bedrock of workplace productivity.

What the Numbers Reveal

Okta’s analysis of anonymized single sign-on access patterns from more than 20,000 companies paints a vivid picture. It tracks enterprise AI tool adoption from June 2022 through June 2026, focusing on platforms for knowledge workers and developers: foundation models, developer tools, enterprise search, collaboration suites, transcription, and creative apps.

The growth race is striking. Setting Anthropic’s net enterprise-account growth to an index of 100, Okta scored OpenAI at 66.9, Google Workspace at 59.8, GitHub at 56.6, and Cursor at 42.4. That means Anthropic widened its lead in the pace of new corporate account acquisition by a significant margin over the four-year period. Notably, Anthropic overtook OpenAI in enterprise customer count in March 2026 and surpassed it in monthly active users a month later.

But growth rates don’t equal market share. By total accounts, Microsoft 365 remains the largest AI-related platform, trailed by Google Workspace. The installed base advantage is real, even if many of those seats aren’t actively using AI features. And that’s the crucial nuance: having a license doesn’t equal adoption, let alone measurable productivity gains.

Microsoft 365: Still the Foundation, But Not a Monopoly

For Windows-centric organizations, Microsoft 365’s dominance isn’t just a statistic. It’s the operational fabric: Outlook for email, Teams for collaboration, SharePoint and OneDrive for files, Entra ID for identity, Intune for endpoint management. AI capabilities embedded inside these tools—whether Copilot in Word, meeting summaries in Teams, or security insights in Defender—arrive in front of users with practically zero installation friction.

That’s the power of incumbency. But the data also reveals a new reality: companies aren’t content to rely solely on Microsoft’s AI ambitions. They’re layering specialist AI tools on top of that foundation. Anthropic’s Claude for deep reasoning and long-context analysis; Cursor for agent-driven coding; OpenAI for flexible developer APIs; dedicated enterprise search platforms—all coexist with a Microsoft 365 backbone.

The multi-platform trend isn’t just about preference. It’s a rational response to a market where no single vendor excels at every task. A coding agent may run circles around a general-purpose chatbot in a codebase, while a transcription specialist beats a suite-wide assistant for meeting accuracy. And as AI moves from simple text generation to agentic actions—calling APIs, updating records, triggering workflows—the risk profile shifts dramatically. Organizations are right to seek best-of-breed in high-stakes domains.

The Multi-Platform Shift: What It Means for You

For Windows administrators, this splintering landscape spells the end of “one AI tool to rule them all” thinking. Your users are likely already using multiple AI services, sometimes through official channels, sometimes not. The Okta finding that 57% of companies use at least two platforms—and that the share of single-provider customers declined from May to June 2026—confirms that shadow AI is no longer a fringe worry; it’s a mainstream operational reality.

The immediate consequences are threefold:

  • Identity governance becomes the linchpin. AI tools don’t just chat anymore; they connect to files, search internal systems, and execute actions on behalf of users. Every new tool introduces another authentication surface. If identity isn’t centralized, you lose visibility into who accesses what and when. Microsoft Entra ID becomes your control plane—if you enforce it.
  • Endpoint security takes on new urgency. An unmanaged Windows PC running a sensitive AI query can leak confidential data, even if your cloud services are locked down. Intune-managed devices with conditional access policies help, but they must be enforced consistently across all endpoints that touch AI services.
  • Agentic permissions demand least-privilege design. When an AI agent uses a service account to read documents or create tickets, that account’s scope becomes critical. Broad permissions granted for convenience during a pilot can turn into a data-exfiltration superhighway at scale. IT must vet these integrations as rigorously as any other privileged account.

The upside? A multi-platform approach can sharpen productivity by matching the right tool to the right job. Developers get Cursor for code, analysts get Claude for long documents, and the rest of the organization gets Microsoft 365 Copilot for everyday tasks. That’s more effective than trying to force a single tool to do everything poorly. The challenge is governance.

How We Got Here: From Experimentation to Operational Necessity

Enterprise AI adoption didn’t start with a grand strategy. It began with curious employees pasting confidential data into public ChatGPT windows. Then came the crackdowns, the education, the rushed procurement of enterprise-grade licenses. For a brief moment, many companies believed they’d pick one winner—probably the vendor that already owned their productivity suite.

But the market moved faster than IT planning cycles. Specialized tools proved their worth in software development (GitHub Copilot, Cursor), advanced analysis (Anthropic), and real-time transcription (meeting assistants). Teams began adopting them independently, often bypassing centralized evaluation. By the time formal vetting caught up, the multi-platform genie was already out of the bottle.

Microsoft knew this would happen. Its strategy has long been to make Windows and Microsoft 365 the “and” platform—the place where work happens, even if a third-party AI model handles the heavy cognitive lift. The new Okta data validates that bet: enterprises are keeping their Microsoft foundation while enthusiastically embracing non-Microsoft AI.

The accelerating tempo is noteworthy. Anthropic’s enterprise accounts surged after January 2026, while OpenAI and Cursor saw growth rates moderate. This suggests that the evaluation criteria have shifted: not just how good a model sounds, but how well it integrates with existing security, how it handles long documents, and whether it can safely perform agentic tasks. For Windows shops, that means the AI governance frameworks designed for one tool must now scale to many.

Your Action Plan: 5 Steps to Manage a Mixed AI Environment

A multi-platform reality doesn’t have to mean chaos. With deliberate steps, you can preserve productivity gains while mitigating risks.

1. Build an AI application inventory. Don’t limit yourself to what procurement knows. Pull data from SSO logs, network filters, expense reports, browser management tools, and even employee surveys. Tag each service as approved, tolerated, or prohibited. You’ll likely uncover surprise subscriptions and shadow-IT usage.

2. Classify use cases by data sensitivity. Not all AI work is equal. Group tasks into tiers: public/low-sensitivity content, internal business information, confidential/proprietary data, highly regulated/personal data, and high-impact actions that alter systems. Define which platforms and permissions are allowed per tier. For example, drafting a marketing blurb might be fine on any approved tool, but analyzing HR records requires the vetted, secured Anthropic deployment.

3. Centralize identity, enforce Entra ID. Every AI service that supports SAML or OIDC should be integrated into your Microsoft Entra ID tenant. This gives you single-pane visibility, conditional access, MFA enforcement, and the ability to revoke access instantly during offboarding. It also prevents the proliferation of password-based accounts that escape governance.

4. Apply least privilege to AI agents. When an AI agent is granted access to a SharePoint folder, Teams channel, or ticketing system, start with read-only. Resist requests for “just in case” broad permissions. Audit agent entitlements regularly, just as you would for human service accounts. Remember: an agent can read files at machine speed—a overprivileged agent is a data breach waiting to happen.

5. Measure adoption and value, not just licenses. It’s easy to buy 1,000 Copilot seats and forget about them. Track active users, workflow improvements, ticket deflection rates, code-review throughput—any metric that ties AI usage to business outcomes. If a tool isn’t delivering, drop it. A lean, effective multi-platform stack beats a bloated one.

Outlook: The New AI Workplace

Expect the enterprise AI landscape to fragment further before it consolidates. New specialist tools will emerge; incumbents will buy or build competing features. For Windows organizations, the strategic imperative is clear: extend your existing identity, endpoint, and data governance frameworks to cover every AI tool that touches your environment. Microsoft Entra, Intune, Purview, and Defender aren’t just Microsoft product names—they become the universal control plane for a multi-vendor AI world.

Anthropic’s momentum shows that enterprise buyers are willing to move fast for tools that deliver clear value. But they’re not ditching Microsoft. The co-existence pattern is likely to deepen. The winners won’t be companies that choose a single AI provider; they’ll be the ones that master the art of securely managing many.