Microsoft Build 2025 is underway, and Microsoft Copilot Studio has already announced many new and exciting features. Behind the scenes, many of these powerful agent capabilities rely on Microsoft Dataverse—the secure, scalable agent platform that extends agents with enterprise data.
As agents evolve to handle more complex, business-critical tasks, they need to be equipped with a new spectrum of tools:
- Data tools to operate human-agent teams for specific business processes
- Knowledge tools to retrieve the right context at the right time
- AI tools to enable complex reasoning and customizable actions
Dataverse is the platform that brings all three together—powering the data, context, and intelligence that agents need to transform business processes with AI. Over the past few months, Dataverse has seen incredible innovation in each of these areas. I’m excited to share a number of new features designed to upgrade agent interaction and performance—whether you’re orchestrating workflows across systems, grounding agents in enterprise knowledge, or unlocking more intelligent and adaptable behavior.
Data tools to operate human-agent teams for specific business processes
Dataverse: the operational database for agents
Dataverse provides the platform for organizations to store, manage, and orchestrate business and operational data across your agent ecosystem. With Dataverse under the hood, Microsoft Copilot Studio makers can deploy agents to handle logic-driven, adaptive tasks at scale while maintaining human oversight where needed.
Dataverse creates a cohesive environment where you can apply generative AI atop your business data and operational data in Dataverse with features like AI-powered Dataverse search, Tools, prompt columns, and Model Context Protocol for seamless, connected operations.
Prompt columns, for instance, bring generative AI directly into Dataverse tables with columns whose values are generated by a prompt. For example, in a product review table with “Review Text” and “Sentiment” columns, you can set “Sentiment” as a prompt column that uses AI to evaluate the review text which returns a value of “Positive” or “Negative” in the column. You can also reference other fields in the row, creating dynamic business logic powered by natural language prompts.
These increasingly sophisticated data-level features lay the groundwork for next-generation, business-aware, agent-first systems. Imagine how you could transform your organization with human-agent teams, with people leading workflows and agents performing tasks.
From an Invoice Processing Agent capturing structured invoice data for later review, to a claims processing system where an autonomous agent handles intake while a chat assistant interacts with users and a person approves the claims, Dataverse is the trusted common data platform for both operational data and business data that these human-agent teams use for business workflows.
To learn more, watch the Microsoft Build 2025 on-demand session: Dataverse for agents.
Dataverse Model Context Protocol server
The Dataverse Model Context Protocol (MCP) server, now in public preview, makes your business data interactive, turning structured Dataverse information into dynamic, queryable knowledge for Copilot Studio agents. For both developers building advanced workflows and makers configuring intelligent experiences, MCP helps make data conversational and usable.
Once connected, the Dataverse MCP server enables four key capabilities:
- Query: Discover available tables, explore schema, and retrieve real-time data via structured or natural language queries
- Knowledge and search: Let agents chat over your data, search knowledge sources, and deliver contextual answers without brittle configurations
- Upload (Create/Update Records): Insert new records or update existing ones in Dataverse, with schema-aware mapping to maintain data integrity
- Generate with grounding prompts: Run custom prompts grounded in real business context (e.g. summarizing a record, evaluating sentiment, or drafting a tailored response)
Exposing your Dataverse environment through an MCP server brings your enterprise data to life by giving your agents the ability to reason across structured data, take informed actions, and generate meaningful outputs while honoring your data model and access controls.
Enterprise data integration across Copilot Studio, Microsoft Fabric, and Microsoft 365 Copilot
Dataverse knowledge in Copilot Studio, now generally available
Agents are only as good as the knowledge they can access, and now AI-powered Dataverse continues to be the backbone of the Dataverse knowledge in Copilot Studio. Dataverse knowledge connects structured and unstructured data from across your organization—including Dynamics 365, Power Platform, and external systems—into a unified, context-rich knowledge network that agents can reason over and act on, including in prompts. New improvements include support of multi-line text and file type columns in Dataverse knowledge.
Use near-real-time data warehousing with Microsoft Fabric
All your data in Dataverse is pre-indexed and ready for near-real-time analytics. Whether it’s critical business data from Dynamics 365, custom Power Apps applications, or configuration and response data from agent interactions, everything is stored securely in Dataverse and instantly available for analysis.
With data constantly updating, Dataverse maintains a near-real-time data warehouse that you can explore using Data Agents in Microsoft Fabric. With just a few clicks in the Power Apps maker portal, you can link Dataverse to Fabric, helping you unlock deep insights on your data.
To make it even easier for Fabric data professionals to get secure access to Dataverse, we introduced Mirrored Dataverse in Fabric. This feature goes to public preview in June 2025.
Dynamics 365 data now available in Microsoft 365 Copilot
Customers can now search and reason over Dynamics 365 data, comprising contacts, accounts, leads, opportunities, and cases, directly within Microsoft 365 Copilot. This new integration brings business and productivity data together, allowing users to glean insights, take action, and stay in the flow without switching contexts.
Previously, users had to rely on a dedicated agent or navigate to Dynamics directly. With this feature (currently in private preview), you can ask questions and complete tasks using both Microsoft Office and CRM system data in one Copilot experience. Join the early access program and help shape what’s next.
Knowledge tools to retrieve the right context at the right time
The knowledge platform in Microsoft Copilot Studio, powered by Dataverse, allows customers to seamlessly use their own data—such as local files and Dataverse tables—as knowledge sources to build intelligent, context-aware agents grounded in their proprietary content. Now it’s easier than ever to unify and operationalize knowledge across your agent ecosystem.
New knowledge sources and connectors
We continue to expand your agents’ reach with new knowledge sources like Snowflake, SAP, Databricks, Confluence (cloud only), OneDrive for Business, and SharePoint Lists. Support for Salesforce, Zendesk, and ServiceNow now includes unstructured content, such as knowledge base articles. For greater functionality, Dataverse and uploaded files now support image extraction, multilingual content, and querying embedded tabular data. Finally, Azure AI Search—a proven solution for information retrieval in Dataverse’s Retrieval Augmented Generation (RAG) architecture—is now generally available as a knowledge source.
Enhanced Power Platform connector Software Development Kit (SDK)
The new enhanced Power Platform connector SDK, now in preview, makes it easier to bring structured external data into Power Apps, Dataverse, and Microsoft Copilot Studio. With the SDK, connectors can expose structured data like full tables and metadata—not just raw APIs. That means that makers can bind tables directly to user interface controls in Power Apps, apply familiar Power Fx functions like Sort and Filter, and ground Copilot agents with business data as a knowledge source, not just an action.
For example, a Databricks connector (public preview in June 2025) built with the enhanced SDK lets makers surface tables in apps and agents automatically. Power Apps understands the schema, and Copilot Studio can answer questions based on the data without manual configuration. Makers and developers can use structured connectors built by software development vendors, or create their own using the SDK.
To learn more, watch the Microsoft Build 2025 on-demand session: Knowledge in Copilot Studio.
AI Tools to enable complex reasoning and customizable actions
Centralized Tools hub in Copilot Studio
The new Tools tab in the left navigation of Copilot Studio gives makers a centralized place to create and manage reusable functionality across all agents in an environment. Using tools, an agent can take actions in external systems (not just read data from them).
There are six tool types in Copilot Studio, rolling out to public preview in June 2025:
- Model Context Protocol, which allows users to connect with existing knowledge servers and data sources directly within Copilot Studio
- Agent flows, which makers can use to automate deterministic and repeatable workflows
- Computer use, which allows agents to navigate and interact with web and desktop applications
- Custom connectors and REST APIs, which both allow makers to connect to third-party systems that aren’t available as prebuilt connectors
- Prompts, which let makers create AI-powered instructions for smarter agents, flows, and apps. With new support for Power Fx expressions, prompts can also perform data transformations like calculations, formatting, or text manipulation
To learn more, watch the Microsoft Build 2025 on-demand session: Tools for agents.
New autonomous agents to put to work faster
Autonomous agents can be a huge boon for your business processes, but building them from scratch takes time…so don’t start from zero. To help makers move faster, we’re introducing three new managed agents, available in preview now in the Create tab of Copilot Studio:
- Document Processor Agent
- Customer Brief Agent
- Lead Manager Agent
The Document Processor Agent is a robust, out-of-the-box solution for automating document workflows. Once installed, it monitors an email inbox for attachments, extracts key information from incoming files, and exports structured data to a target system. When needed, it seamlessly requests human validation, routing documents to assigned reviewers and tracking progress through an integrated monitoring app. Notifications are sent via Teams or Outlook, and validators can view, correct, or approve extracted content in just a few clicks.
Unlike traditional document automation—where each solution generally starts from scratch—this agent comes prebuilt and requires minimal setup. Using prompts with multimodal input, there is no model training required and no need to build separate validation apps or flows. Going from install to action can take minutes, not hours.
To learn more about the Document Processor agent, watch the Microsoft Build 2025 on-demand session: Agents in action: Document processing 2.0.
The other new templates help automate other mission-critical tasks:
- The Customer Brief Agent pulls from your business data to generate timely, relevant executive briefs before client meetings.
- The Lead Manager Agent acts as a top-of-funnel assistant, autonomously processing and responding to inbound leads.
All three agents are designed to help you get started quickly and scale faster. Learn more at Microsoft Build 2025: Build autonomous agents in Copilot Studio.
All these Dataverse updates represent a major leap forward in how makers and developers can build, scale, and manage intelligent agents with Dataverse at the core. From deeper integrations to smarter tools and more powerful out-of-the-box agent templates, Copilot Studio is the go-to platform for business-ready AI.
To dive deeper into these new capabilities, join us for some of our relevant sessions at Microsoft Build 2025:
- Building safe and reliable Microsoft Copilot Studio agents
- GenAI for Enterprise: Intelligent Apps and Agents with Dataverse & MCP
- Enhancing Agent Management Controls for AI-Ready Enterprises
- Unlock your Dataverse data with secure and powerful search
- Agents in action: Document processing 2.0
- Tools for agents
- Dataverse for agents
- Knowledge in Copilot Studio
