Claude’s Next Enterprise Battle Is Not Models: It’s the Agent Control Plane

New data from VentureBeat Pulse shows Microsoft and OpenAI firmly leading enterprise agent orchestration. However, Anthropic’s first measurable foothold points to a much larger architectural war over who controls the critical infrastructure where AI agents actually run.

For the last two years, the enterprise AI race has been framed almost entirely as a model war: OpenAI’s GPT series versus Anthropic’s Claude versus Google’s Gemini. But the next strategic battleground isn't about which model answers a prompt best. It’s about who controls the layer where agents plan, call tools, access data, manage workflows, and prove to security teams that they haven't gone rogue.

The Shift to Native Orchestration

New VB Pulse survey data—which tracks the preferences of verified, technical decision-makers at enterprises—suggests this new category is rapidly taking shape.

Primary Enterprise Agent Orchestration Platform Adoption (Jan–Feb 2026)

Platform	January Adoption	February Adoption	Trend
Microsoft (Copilot Studio / Azure AI Studio)	35.7%	38.6%	▲ Rising
OpenAI (Assistants & Responses API)	23.2%	25.7%	▲ Rising
Anthropic (Tool Use & Workflows)	0.0%	5.7%	▲ Emerging
Independent Frameworks (LangChain / LangGraph)	5.4%	1.4%	▼ Falling
External/Abstracted Orchestration	8.9%	2.9%	▼ Falling

While Anthropic's 5.7% foothold represents a small cohort (four out of 70 respondents), it marks a critical strategic milestone: the first sign of Claude usage migrating from a plug-and-play model layer into native orchestration.

Enterprises are no longer just choosing chatbots. They are deciding where the live, operational machinery of AI work will sit: inside Microsoft’s stack, OpenAI’s API layer, Anthropic’s managed runtime, or a hybrid mix of all of them.

“This is the convergence moment for enterprise AI,” says Tom Findling, CEO and cofounder of AI cybersecurity startup Conifers. “Models and agent frameworks have matured enough together that enterprises are shifting focus beyond model quality to the control plane around it. Competitive advantage is moving toward platforms that can orchestrate agents, leverage enterprise context, and provide governance and auditability.”

Why an Agent Runtime is Sticky

A model is relatively easy to swap. A company can route one workload to Claude, another to GPT, and another to an open-source model based on cost and risk. In fact, VB Pulse data shows that a multi-model approach is the enterprise consensus.

An agent runtime, however, is infrastructure. Once a company’s workflows, tool permissions, credentials, audit logs, and sandboxed execution environments live inside a specific provider’s ecosystem, switching costs skyrocket.

Anthropic is well aware of this stickiness. Their Claude Managed Agents public beta decouples the model from the surrounding agent machinery (the session, the harness, and the sandbox). Anthropic is actively positioning itself to host a secure environment where agents operate over long-running workflows.

Security Over Flexibility

As agents gain the power to send emails, query databases, and modify files, their potential blast radius expands. Consequently, enterprise buying criteria are shifting dramatically from optionality toward strict governance.

Top Selection Criteria for Agent Orchestration

Security & Permissions: 37.1% (Ranked #1)
Control Over Agent Execution: 22.9% (Up from 17.9% in January)
Flexibility Across Models/Tools: 25.7% (Down from 35.7% in January)

Ev Kontsevoy, CEO of Teleport, warns that orchestration without identity multiplies chaos:

“Without identity, you don’t know what an agent can access, what it actually did, or how to revoke its access when it operates outside policy. A unified identity layer is a prerequisite to deploying agents.”

Syam Nair, Chief Product Officer at NetApp, echoes this sentiment, stressing that data and metadata visibility are crucial for enforcing clear guardrails at scale across security, storage, and data science teams.

The Walled Garden vs. The Hybrid Reality

Microsoft holds a massive enterprise distribution edge because Copilot Studio natively integrates with tools companies already use: Microsoft 365, Teams, Azure, and Entra ID. David Weston, CVP of AI Security at Microsoft, notes that customers want to "bring order to complexity" through a single control plane like Agent 365 to observe and secure agents across ecosystems.

Yet, despite the convenience of provider-native stacks, enterprises are deeply wary of vendor lock-in.

VB Pulse Risk Data: Vendor Lock-in Concerns
Jan 2026: 23.2% 
Feb 2026: 25.7% ▲ (The only concern to increase month-over-month)

Because of this, a hybrid control plane—combining provider-native orchestration with external guardrails—remains the leading expected architecture, holding steady at 35% to 36% of the market.

Dr. Rania Khalaf, Chief AI Officer at WSO2, emphasizes that the industry is transitioning from LLMOps to "Agent Ops."

“A guardrail on an LLM call can catch a hallucination, but it won't catch an agent thrashing in an unbreakable, costly loop. Pulling guardrails, evals, and agent identity out of the core logic allows them to be configured per deployment without fragmenting governance as teams choose different models.”

The Battle Ahead

To counter lock-in fears, Anthropic introduced the Model Context Protocol (MCP), an open-source standard for connecting AI to data and tools. However, openness at the protocol layer doesn't completely eliminate runtime dependency. Companies using MCP may still find themselves tied to Anthropic’s managed sessions and sandboxes.

Meanwhile, startups like BAND (led by CEO Arick Goomanovsky) are betting on a "cross-vendor collaboration layer" that allows agents from Microsoft, OpenAI, and Anthropic to interact seamlessly as a unified workforce.

The agent market is beginning to look less like software-as-a-service and more like cloud infrastructure. The winners won't just have the smartest models; they will provide the most robust identity integration, audit logging, and observability.

Anthropic doesn’t need to dethrone Microsoft tomorrow. If it can convince its rapidly growing base of Claude enterprise power-users to let it manage the surrounding operational machinery, Claude ceases to be just a model in a portfolio—it becomes the infrastructure where work gets done.