EP 529: Microsoft Build Updates: 5 new Copilot AI updates and how to use them
🎯 Summary
Podcast Summary: EP 529: Microsoft Build Updates: 5 new Copilot AI updates and how to use them
This episode of the Everyday AI Show, hosted by Jordan Wilson, provides a focused breakdown of the most impactful AI updates announced at the Microsoft Build 2025 conference, specifically highlighting five key advancements related to Microsoft Copilot that business leaders should understand. The host filters through the massive volume of announcements (an 80-page book of news) to deliver actionable insights on how these new tools can transform business operations.
1. Focus Area
The primary focus is on Enterprise AI and Agentic Workflows stemming from Microsoft Build 2025. The discussion centers on five major updates to the Copilot ecosystem, emphasizing the shift from simple assistants to autonomous, multi-agent systems capable of complex, cross-application tasks.
2. Key Technical Insights
- GitHub Copilot as an Autonomous Agent: GitHub Copilot has evolved beyond a coding assistant to an autonomous coding partner embedded in VS Code, capable of iterative refinement, testing, and accepting multimodal inputs like screenshots/mockups. It now supports the Model Context Protocol (MCP) popularized by Anthropic.
- Agent Foundry and Interoperability: The generally available Azure AI Agent Foundry allows users to design and scale enterprise agents using thousands of models (including Grok 3, Llama, and GPT models). Crucially, it supports both Google’s Agent-to-Agent (A2A) protocol and Anthropic’s MCP, enabling agents across different platforms to communicate.
- Multi-Agent Orchestration: Copilot Studio now natively supports multi-agent orchestration, allowing specialized agents to discover, negotiate, and collaborate on tasks securely, leveraging agentic memory and automated validation tools.
3. Business/Investment Angle
- Copilot Tuning for Customization: The introduction of low-code Copilot Tuning allows organizations to fine-tune the core Copilot model using internal data and workflows (e.g., brand voice, output structure) without needing data science expertise, operating securely within the M365 boundary.
- High Barrier to Entry for Tuning: Currently, Copilot Tuning is restricted to large enterprises (requiring 5,000 Copilot licenses or an account team contact), which the host views as a potential misstep, limiting adoption by agile mid-market companies.
- Automation Potential via Computer Use: The “Computer Use in Copilot” feature promises significant ROI by automating highly repetitive, mundane desktop tasks like data entry, invoice processing, and cross-application navigation, acting like a human user via natural language prompts.
4. Notable Companies/People
- Microsoft: The central entity, driving these updates across GitHub, Azure, and M365 Copilot.
- Anthropic: Mentioned for popularizing the Model Context Protocol (MCP), which is now being adopted widely.
- Google: Mentioned for its Agent-to-Agent (A2A) protocol, which Microsoft is now supporting for interoperability.
- Elon Musk (xAI): Mentioned in the context of Grok models being available within the Agent Foundry.
- Jordan Wilson (Host): Positioned as an expert consultant helping companies navigate Gen AI strategy and ROI, referencing partnerships with Adobe, Nvidia, and others.
5. Future Implications
The industry is rapidly moving toward agentic computing, where specialized AI agents collaborate autonomously to execute complex business processes. Interoperability standards like MCP and A2A are becoming critical infrastructure, allowing agents to communicate across vendor ecosystems. The future of work involves automating entire workflows rather than just single tasks, though the host cautions that multi-agent environments require extreme care regarding governance and alignment to prevent compounding errors.
6. Target Audience
This episode is highly valuable for Enterprise IT Leaders, AI Strategists, Software Developers, and Business Leaders who use or plan to implement Microsoft 365 Copilot. It provides a curated, professional summary of complex technical releases, focusing on practical application and strategic implications.
Comprehensive Narrative Summary
The podcast episode serves as a vital digest of Microsoft Build 2025, focusing exclusively on five (plus a bonus sixth) transformative updates to the Copilot ecosystem designed to elevate AI from a productivity tool to an autonomous business partner.
The narrative begins by establishing the overwhelming volume of announcements and immediately pivots to the host’s curated list of the most important changes. Update 1 details the evolution of GitHub Copilot into an autonomous coding agent that accepts multimodal inputs and supports the MCP standard, positioning it as a full programming suite rather than just an assistant.
Update 2, Copilot Tuning, is presented as a potential game-changer for enterprise differentiation. This low-code feature allows organizations to fine-tune the base Copilot model using proprietary data for specific tasks, tone, and structure, all while ensuring data security (Microsoft confirms customer data is not used for foundation model training). However, the host critiques the high entry barrier (5,000 licenses), suggesting it favors slow-moving giants over agile mid-market firms.
The discussion then moves into the realm of Agentic AI. Update 3, Agent Foundry, is an Azure-powered playground for building, deploying, and scaling enterprise agents, supporting thousands of models. Its significance lies in its adoption of both Google’s A2A and Anthropic’s MCP protocols, signaling a necessary move toward open standards for agent communication.
This leads directly into Update 4, Multi-Agent Orchestration in Copilot Studio. This feature allows multiple specialized agents to
🏢 Companies Mentioned
đź’¬ Key Insights
"That's why this is huge. So, MCP is supported by default inside of Windows, native MCP support. And I can't emphasize enough how big that is for the future of multi-system agentic AI, right?"
"Like I said, the MCP protocol was designed by Anthropic. So, it's not just in the AI Foundry. It's not just in Copilot Studio with the multi-agents. It is literally supported almost everywhere, including GitHub Copilot, Azure Dynamics, Copilot Studio, Microsoft 365 apps, and even Windows 11."
"MCP native support. So, here's why this is big. It's now Model Context Protocol for agents' access to tools, system, and enterprise data."
"But you'll see here, I mean, it's a live computer-using agent, which right now, you know, you have to—I mean, Operator from OpenAI is the only other mainstream one that's at least easy to use. Anthropic has their computer-using agent, but it's not web-based, and it's actually extremely clunky and very hard to use."
"computer use in Copilot. So right now, this enables agents to automate websites and desktop applications via natural language. So agents can interact like humans, clicking buttons, typing, and navigating interfaces."
"throw out 'human in the loop'—I hate that—you need to have expertise in the loop with the right humans on the oversight as well."