Microsoft Build 2025 happened last week, and as expected, AI dominated the agenda. Three days of announcements, demos, and roadmap previews – most of which won’t affect your enterprise planning.

I’ve distilled the relevant parts. Here’s what matters.

Copilot Evolution

Microsoft’s AI assistant strategy continued to expand:

Copilot Studio enhancements. The low-code tool for building custom Copilots got significant upgrades. Better integration with enterprise data sources, improved action capabilities, and more sophisticated conversation flows.

This matters because it lowers the barrier to building AI assistants tailored to specific workflows. Organisations can create Copilots for IT help desk, HR queries, sales support, and other domain-specific needs without extensive development.

Copilot for Microsoft 365 improvements. Better performance, faster responses, more accurate document summarisation. Incremental improvements that address common user complaints.

Copilot agents. The ability to define autonomous agents that can take multi-step actions – schedule meetings, update records, trigger workflows. This is powerful but requires careful governance.

My take: Copilot is evolving from assistant to platform. The custom Copilot capability is the most significant announcement for enterprises. The challenge is governance – agents that can take actions need oversight.

Azure AI Platform Updates

GPT-5 integration. Azure OpenAI now includes GPT-5 access, though initially in limited preview. The capability jump from GPT-4 is meaningful for complex reasoning tasks.

Improved fine-tuning. Easier processes for customising models with organisation-specific data. Still requires expertise but less than before.

Better RAG (Retrieval Augmented Generation). Azure AI Search integration with Azure OpenAI is tighter and performs better. Document grounding is more reliable.

Cost optimisation features. New options to route requests to appropriate model tiers based on complexity. Simple queries use cheaper models; complex ones use capable models.

My take: Azure AI continues maturing as an enterprise platform. The cost optimisation features are particularly welcome – token costs at scale have been a pain point.

Developer Tooling

GitHub Copilot Enterprise. Expanded capabilities for enterprise development teams, including better code review, documentation generation, and security scanning.

Visual Studio integration. Deeper AI assistance throughout the development lifecycle, not just code completion.

Power Platform AI. More AI capabilities in Power Apps, Power Automate, and Power BI. The low-code platforms are becoming AI-native.

My take: For organisations with significant development teams, the GitHub Copilot improvements are meaningful. The Power Platform AI features make citizen development more capable but also riskier – governance is important.

Security and Compliance

Microsoft Secure Future Initiative updates. Enhanced security for AI workloads, better audit capabilities, and improved threat detection for AI systems.

Compliance certifications. Additional regulatory compliance certifications for Azure AI services, including Australia-specific requirements.

Responsible AI tooling. Better tools for testing AI systems for bias, safety, and reliability before deployment.

My take: These announcements won’t make headlines but matter for enterprise adoption. Security and compliance capabilities have been a genuine barrier; these improvements help.

What’s Missing

Pricing clarity. Many new features don’t have clear pricing yet. Budget planning remains difficult.

Timeline commitments. “Preview” and “coming soon” appeared frequently. Enterprises need to know when features will be production-ready.

Migration paths. As Copilot capabilities expand, guidance on migrating existing AI investments is limited.

The Competitive Context

Microsoft is maintaining its lead in enterprise AI through:

• Tight integration with Microsoft 365 (which most enterprises use)
• OpenAI partnership providing access to frontier models
• Enterprise-grade security and compliance
• Extensive partner ecosystem

Google and AWS are catching up on capability but struggle to match Microsoft’s integration advantage for Microsoft-native organisations.

What This Means for Enterprise Planning

If you’re already Microsoft-committed: These announcements reinforce that bet. Evaluate Copilot Studio for custom AI assistants. Explore GPT-5 when available.

If you’re evaluating platforms: Microsoft remains the safe choice for enterprises invested in Microsoft 365. The integration advantages compound over time.

If you’re multi-cloud: The AI-specific announcements are less relevant – similar capabilities exist elsewhere. But Copilot integration with Microsoft 365 is unique.

My Recommendations

Based on Build 2025:

1. Evaluate Copilot Studio if you need domain-specific AI assistants. The capability is maturing.

2. Plan for Copilot agents but establish governance first. Agents that can take actions need human oversight.

3. Test GPT-5 when available. The capability improvements may enable use cases that weren’t viable before.

4. Leverage cost optimisation. If you’re spending significantly on Azure OpenAI, the new routing features can reduce costs.

5. Update your security assessment. If security concerns have blocked AI adoption, review whether the new features address your requirements.

The Bigger Picture

Build 2025 wasn’t revolutionary. It was evolution – continued improvement of existing capabilities, tighter integration, and incremental progress on enterprise concerns.

That’s actually appropriate for where the market is. Enterprises don’t need revolutionary announcements; they need reliable, governable, integrated AI capabilities. Microsoft is delivering that.

The platform wars continue, but for most enterprises already invested in Microsoft, the question isn’t “should we switch?” It’s “how do we get more value from what we have?”

Build 2025 provided some answers to that question.