Microsoft Build 2025 Digest
Key AI/ML/Data/Analytics announcement digest
TLDR: agents, agents
Azure
Agentic web: Azure AI Foundry Models and new tools for model evaluation; Azure AI Foundry Agent Service (supports Agent2Agent (A2A) and Model Context Protocol (MCP))/Azure AI Foundry Observability; Microsoft Entra Agent ID (unique identities in an Entra directory, Purview governance); Supporting Model Context Protocol (MCP) on authorization specification, MCP server registry service; NLWeb (makes it easy for websites to provide a conversational interface, every NLWeb endpoint is also an MCP server)
Azure SRE Agent: helps respond to incidents quickly and effectively and manages production environments
Microsoft Discovery: research and development (R&D) agentic platform: from advanced knowledge reasoning and hypothesis formulation to experimental simulation and iterative learning
Model router for Azure AI Foundry is a deployable AI chat model that is trained to select the best large language model (LLM) to respond to a given prompt in real time.
Agentic retrieval: uses a new query architecture that incorporates user conversation history and an Azure OpenAI model to plan, retrieve and synthesize queries.
Data agents in Microsoft Fabric: virtual business analysts, like Databricks’ Genie, hich you can bring right into Copilot Studio and deploy across Microsoft Teams and Microsoft 365 Copilot.
Announcing the availability of Azure Databricks connector in Azure AI Foundry: create AI agents that leverage their valuable enterprise data stored in Microsoft Azure Databricks
SQL Server 2025: vector data type allows hybrid AI vector searches, vector functions, AI model management provides options for deploying AI models either on-premises or in the cloud, with compatibility for Azure OpenAI, OpenAI endpoints, and Ollama.
Digital twin builder in Microsoft Fabric: real-time intelligence
AI-powered transformations on OneLake: AI-powered transformations such as summarization, translation and document classification
Edge
Windows AI Foundry (run open source LLMs via Foundry Local)
Native support for Model Context Protocol (MCP) on Windows
Developer
GitHub Copilot coding agent, GitHub Models
open source GitHub Copilot Chat in Visual Studio Code
Microsoft App Modernization Guidance, 2, AWS
Project Amelie: autonomous agent from Azure AI Foundry Labs with Microsoft Research, which can build complete machine learning pipelines from a single prompt.
M365
Microsoft 365 Agents Toolkit for Visual Studio, Microsoft 365 Agents SDK enables building multichannel agents
Teams AI library: Teams-centric interface for integrating GPT-based language models and user intent engines
Microsoft 365 Copilot APIs: securely access Microsoft 365 Copilot capabilities in your own applications and custom engine agents
Microsoft Copilot Studio and Microsoft 365 Copilot support multiple agent systems (agents built using Copilot Studio, Azure AI Foundry Agent Service or the Microsoft 365 Agents SDK) through Agent2Agent (A2A), tool access with Model Context Protocol (MCP), computer use tool to perform enterprise tasks on user interfaces across both desktop and web apps, Code Interpreter allow agents to write and run Python code to perform complex tasks, storage is powered by Dataverse and Dataverse search
Agent Store, 2: Build, publish, and discover agents in Microsoft 365 Copilot
Microsoft 365 Copilot Tuning (use own company data, workflows and processes to train Copilot), multi-agent orchestration in Copilot Studio
