Microsoft opened its Build 2026 developer conference in San Francisco on June 2 with a keynote that made the company’s strategic intent unmistakable: Windows is now an agent platform, and Microsoft is no longer content to be a distribution channel for other companies’ AI models.

The announcements at Build 2026 are the most consequential since Satya Nadella bet the company on AI in 2023. The difference this time is that Microsoft is building the infrastructure itself.

Project Polaris: Microsoft’s Own AI Model

The headline announcement is Project Polaris — Microsoft’s in-house AI coding model, built to replace GPT-4 Turbo as the default reasoning engine for GitHub Copilot starting in August 2026.

Polaris uses a mixture-of-experts architecture with specialized sub-modules for different programming languages. According to Microsoft’s benchmarks presented at Build, Polaris outperforms GPT-4 Turbo on HumanEval and MBPP across most languages, with particularly strong results in lower-resource languages like Rust and Haskell.

For existing GitHub Copilot subscribers, the migration to Polaris is automatic. Teams that want to stay on GPT-4 Turbo have a three-month fallback window to opt out before Polaris becomes the sole default.

The strategic implication is significant. Microsoft has been paying OpenAI for model access since its 2019 investment. Deploying a homegrown model for its highest-volume developer product reduces that dependency — and gives Microsoft full control over the capability roadmap for Copilot going forward.

Windows Is Now an Agent Runtime

Beyond Project Polaris, Microsoft formally repositioned Windows itself at Build 2026.

Windows Agent Framework goes open source. The Windows Agent Framework (WAF) — previously available only to select enterprise partners — was released under an MIT license at Build 2026. The framework handles agent lifecycle management: registration, startup, shutdown, memory, and cross-agent communication via a gRPC-based bus. Developers can now build on it, fork it, or embed it into their own products.

Agent Mode is the default for Microsoft 365 Copilot. Microsoft announced that Agent Mode is now the default interface for Microsoft 365 Copilot across all plans. Rather than prompting Copilot as a chat assistant, users interact with it through a task-oriented interface where the agent can take multi-step actions across Outlook, Teams, Word, and Excel.

Azure Agent Mesh. A new control plane that federates agent execution across on-premises Windows servers, Cloud PCs, and Azure Arc-enabled edge devices. Azure Agent Mesh gives IT administrators a single view of what agents are running, where, and what they’re doing — regardless of where the compute lives. General availability is scheduled for Q4 2026.

Copilot Workspace exits beta. Microsoft’s agentic development environment — where teams can spin up multi-agent coding projects with shared context, parallel execution, and coordinated output — is now generally available.

What Wasn’t Announced

No Windows 12. Microsoft confirmed before Build that the conference would be focused entirely on developer tooling and agent infrastructure. Hardware announcements and operating system updates are off the agenda.

What This Means for Business

The Build 2026 announcements carry direct implications for how enterprises think about AI agent strategy.

Microsoft is moving AI agents into the OS layer. The combination of Windows Agent Framework, Azure Agent Mesh, and the Windows Agent Store creates a native runtime for enterprise agents that didn’t exist eighteen months ago. For businesses that run on Microsoft infrastructure — and most do — this dramatically lowers the cost and complexity of deploying agents in production.

Developer tooling is consolidating. Project Polaris replacing GPT-4 in GitHub Copilot is not just a vendor swap. It signals that Microsoft intends to own the full stack for developer AI assistance — the IDE, the model, the deployment infrastructure, and the governance layer. Teams evaluating developer AI tools should factor in that Microsoft’s offering is becoming vertically integrated in a way competitor products are not.

Agent governance is getting standardized. The AgentPolicy API in Windows Agent Framework lets IT administrators define granular controls on what each agent can access. This addresses the most common reason enterprise security teams block AI agent deployments: no governance primitive at the OS level. With WAF open-sourced and the policy API embedded, governance becomes a configuration problem rather than an architecture problem.

Azure Agent Mesh solves the enterprise deployment gap. One of the real blockers to enterprise AI agent adoption hasn’t been the models — it’s been hybrid infrastructure. Most large businesses have on-premises systems they can’t move to the cloud, legacy Windows servers that run critical workloads, and edge compute that isn’t Azure-managed. Azure Agent Mesh treats all of these as valid execution environments, which opens agent deployments to a broader set of enterprise workflows than cloud-only approaches can reach.

The Bigger Pattern

Microsoft’s Build 2026 is a declaration that it will control its own AI stack. Project Polaris reduces OpenAI dependency. Windows Agent Framework creates a proprietary agent runtime. Agent Mode as default pushes enterprise users toward agentic workflows whether they opted in or not.

This is the infrastructure play that comes after the experimentation phase. The businesses that start deploying agents on this infrastructure in 2026 will have compounding advantages over those who wait for the ecosystem to settle.

---

Interested in deploying AI agents inside your existing Microsoft infrastructure? Enterprise DNA’s Omni Ops team works with businesses to identify, build, and govern agent-based workflows — using the platforms you’re already running.