Dataverse at Build: The Agent Platform Powering the Future of Agentic AI

Here are some of the Dataverse announcements from Microsoft Build. With so many new capabilities, here are all our key sessions in one place. To learn more about these exciting updates, read the full Dataverse Build blog. 

Microsoft Dataverse is 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 models, knowledge, tools, and orchestration.

Dataverse is the platform that brings all these capabilities 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 focus areas as we continue to deliver a robust, end-to-end platform for data integration enabling seamless connectivity across your data estate. This means all your data (any shape, any size) can be transformed into Knowledge for agents built in Copilot Studio. We’ve also made strides as the most trusted platform to scale your agentic AI, agents, and apps.

You can see this in action in the “GenAI for Enterprise” session, where we showed how data is transitioning from a static set of information into knowledge that is deeply infused in agents to better understand your relevant business context. Dataverse serves as a unified agent database (with a managed vector index) to enable makers to build apps and agentic experiences on your enterprise systems (LOB systems, APIs, and custom connectors) with its own Model Context Protocol (MCP) server. Watch the full session:  

Key takeaways:  

The Dataverse MCP server is the backbone for teams of agents and solves common data problems, indexing knowledge for context and powers AI actions as part of the end-to-end business process. MCP breaks barriers for your enterprise systems and business processes. By building on Dataverse, the enterprise data platform that powers Copilot Studio, your data is integrated across the growing ecosystem of AI apps and agents. To take a deep dive into this topic, see the “Build your next-gen agents using Dataverse MCP server” session listed below. 

The Common Knowledge Graph in Copilot Studio connects your agent to your enterprise systems with new enhanced connectors. The simplifies RAG so makers spend less time on indexing and integration to focus on building agents. The self-service Developer SDK to build enhanced Power Platform connectors enables makers to build agents in Copilot Studio grounded in enterprise knowledge. To take a deep dive into this topic, see the “Knowledge in Copilot Studio” session listed below.  

Document Processor agent is an out of the box solution to automate and optimize document processing workflows by extracting valuable information. For example, the agent can be triggered by an inbound email, use built in actions to process the information on the email, and store that data in Dataverse, as the enterprise data platform, so all your agents have that shared context. To take a deep dive into this topic, see the “Agents in action: Document processing 2.0” session listed below. 

Knowledge Sessions 

Knowledge in Copilot Studio

We explore how to effectively leverage your enterprise knowledge in Microsoft Copilot Studio to enhance your agents with more relevant and contextual answers. You’ll discover the diverse types of knowledge available and learn how to optimize them to meet your specific business needs. Use knowledge from uploading files, using Dataverse tables and RAG search over connected data sources including SQL, Salesforce, Snowflake, and more. 

Key takeaways:  

Now it’s easier than ever to unify and operationalize knowledge across your agent ecosystem. The Common Knowledge Graph in Copilot Studio connects your agent to your enterprise systems with new enhanced connectors and simplifies RAG so makers spend less time on indexing and integration to focus on building agents. 

Pattern 1: Ingests data to create vector embeddings. Examples include Files, OneDrive, Dataverse, Salesforce & ServiceNow (unstructured data), Confluence, and Dynamics 365 (F&O v-tables) 

Pattern 2: Indexes metadata to query. Examples include Azure SQL, ServiceNow, Salesforce, Zendesk, SharePoint List, Snowflake, Databricks, SAP (Hana), and Oracle. 

Pattern 3: Read operations on Power Platform connectors. Examples include Azure AI Search and 1.4K+ Power Platform connectors 

Customers including CSX, ESET, and Encino Energy are using enterprise Knowledge to build (or have plans to build) agents with Copilot Studio to improve business processes.  

Build your next-gen agents using Dataverse MCP Server

Connect agents to your enterprise data seamlessly using Model Context Protocol (MCP). The Dataverse MCP server can integrate with any AI agent, whether in Microsoft Copilot Studio, Claude, or even directly within Visual Studio Code. This unlocks key insights through knowledge search and retrieval to empower agents to take meaningful actions. 

Key takeaways:  

Model Context Protocol (MCP) is an open protocol that standardizes how apps provide context to LLMs making MCP the “USB-C port” for AI Apps. To ensure a secure framework applying the latest enterprise security including authentication and DLP policies, Microsoft will now certify MCP servers.  

We launched the Dynamics 365 MCP Servers (Sales, Supply Chain, Service, Finance, and Business Central) and the Dataverse MCP server which brings your data to life by making data conversational as agent ready intelligence.  

Query: Discover available tables and run queries to retrieve data. 

Knowledge & Search: Natural language chat over your business data. 

Upload (create/update records): Insert new records mapped to your schema and modify fields of existing records. 

Generate with grounding prompts: Run a prompt grounded in business context – like summarizing a record or drafting a response. 

Microsoft partners with Litera and Docusign to build MCP connectivity so agents can do more.  

Litera MCP Server is designed to advance law firm intelligence and orchestrates across multiple domains with precision and control. For example, incoming resumes can now be augmented by querying against firm wide relationship history in real time.

Docusign MCP Server enables intelligent agreement workflows. For example, in a new employee onboarding scenario, managers can easily track offer letters, send reminders to candidates, understand agreement details, and kickoff new employee onboarding.  

Unlock your Dataverse data with secure and powerful search

Explore how Dataverse insights are easily discoverable across multiple surfaces—Global Search and the Data Exploration agent in Model-Driven Apps, Dataverse Knowledge in Copilot Studio, and MCP clients like Claude—using AI-powered indexes. 

Key takeaways:  

Smarter Global Search: Personalized relevance ranking and semantic search help users find the right Dataverse record faster across your business apps. 

Natural Language filtering and visualization: Filter and visualize data using natural language through the Data Exploration agent in your business apps. 

Ground your agents with Dataverse knowledge: Converse with your business data (including multi-line text and file columns) using agents built with Copilot Studio or in any MCP Clients powered by Dataverse MCP Server.   

Agents with real-time knowledge from Databricks

Transform Databricks data to knowledge for agents by connecting to knowledge sources with the latest enhanced Databricks Power Platform connector. To create this agent, we start in Copilot Studio and go through the step-by-step maker journey to build and deploy an agent based on an existing business workflow.

Key takeaways:   

Power Platform connectors ground agents in your enterprise knowledge resulting in more accurate, context relevant responses. We partnered with Databricks to bring Databricks data in a structured format for agents to read with the enhanced Databricks Power Platform connector. This new class of enhanced power platform connectors makes your agent’s output more predictable, reliable, and improves execution speed.   

For ISV interested in building connectivity to your own data source, we make that easy with the self-service Developer SDK to build enhanced Power Platform connectors for agents.  

Tools and Models Sessions 

Tools for Agents

Makers can enhance the quality of the agent’s response in Copilot Studio by using Tools including and Prompts and agent flows. 

Key takeaways:  

Tools in Copilot Studio is a one stop shop to extend the functionality of an agent. Tools are defined by the following conditions:  

Extend an agent’s capabilities to enact change on an external system 

Capabilities that can be used across multiple agents 

Enabled via a click to install workflow 

There are a wide variety of Tool types that can be used across a range of scenarios which you can reuse when building agents: 

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. 

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. 

Code Interpreter will allow agents to write and run Python code to perform complex tasks such as calculating math, advanced analytics, and generating data visualizations.  

AI in action: Document Processing 2.0

Copilot Studio revolutionizes document processing and content generation. Pre-trained for seamless content extraction, validation, and workflow automation, Document Processing agent enables effortless, end-to-end document handling through connectors.   

Key takeaways:  

80% of enterprise processes rely on documents (invoices, statements, applications, records, etc.). Makers can simplify email, document, and image processing with the Document Processor agent, which is an out of the box agent that understands, extracts, and takes action on documents (with inbuilt Multimodal Prompts and Validation Station). This happens in an autonomous workflow, supervised by humans:  

Collect: Agent monitors event triggers (ie. incoming emails) and scans data sources to collect documents.  

Classify: Agent classifies the document type. 

Extract: Agent reads, interprets and transforms content with GenAI.  

Validate: Agent validates extracted data (based on maker defined business rules and/or human input). 

Intergate: Agent connects to target systems, knowledge, repositories, and apps to process the data. 

We introduced new models including Custom Document Processing 4.0, Healthcare, Contracts, and Tax forms.

To help makers further tailor their agent so they can behave in a predictable way, we improved the Prompts with Prompt Library (60+ templates), Prompt Monitoring, Power Fx Expressions, and Connector Knowledge. 

Orchestration Session 

Dataverse for agents

Dataverse helps you build the future of innovation atop your business data. In this session, we share our customers’ journeys of building team of agents and copilot-powered app development in a friendly and low-code way, backed by Dataverse. This session covers AI-powered search in agents and apps, copilot-assisted data modeling and business logic creation, and how Dataverse powers the operations of agents created in Copilot Studio. 

Key takeaways:  

Dataverse unifies common data for your apps and stores operational data for your business apps. We are evolving Dataverse to enable makers to easily apply generative AI to your existing data and store operational and business data for your agents. Makers can build intelligent agentic apps so that every person is augmented by AI that understands their specific business context:  

Makers: Data modeling with Plan Designer in Power Apps  

Markers: Describing business logic (via prompts) as a prompt column in Dataverse 

Business users: Exploring data with Data Exploration agent in apps  

In AI first organizations, teams of AI agents work together with humans to accomplish business workflows with Dataverse at the center for intelligent, highly coordinated agents. Three key features that illustrate building team of agents include:  

Dataverse Knowledge which leverages out-of-the-box RAG to transform your business data as real-time Knowledge for your agents.  

Integrate reusable Functions as actions in Copilot Studio, to execute specific tasks in agents. 

Storing operational data for teams of autonomous agents to provide a shared context for complex multistep business workflows.  

Demos from Velrada and Litera to demonstrate how teams of agents and humans can seamlessly work together using Copilot Studio, Dataverse, and the Microsoft ecosystem.  

Agent Management Sessions

Enhancing Agent Management Controls for AI-Ready Enterprises

Deep dive into agent management controls. We’ll focus on enterprise-grade security, maintaining healthy and seamless operations, and governing at scale. Watch to gain insights into best practices, tools, and strategies to ensure their organization is AI-ready. Discover how to leverage a robust management suite to enhance your development processes and secure your enterprise environment. 

Key takeaways:  

Agent builders can leverage ALM and DevOps practices when building and deploying agents 

Admins now have a range of capabilities in MCS and PPAC to securely govern agent development in their organizations at scale. Makers can use new capabilities in PPAC Security Hub to enhance your organization’s security posture.  

Leverage Microsoft Information Protection (MIP) labels to protect data Dataverse and Microsoft Copilot Studio.  

Optimize agent costs using the new Copilot Studio Messages Consumption Estimator and cost management tools in PPAC. 

Mitigate the risk of outages, and run disaster recovery drills with cross-region self-service disaster recovery 

Building safe and reliable Microsoft Copilot Studio agents

With the fully managed suite of capabilities for Power Platform, admins and makers alike are equipped with the necessary tools to ensure that Copilot Studio agents are protected and healthy. Tune in to learn more about the latest enhancements and upcoming plans for a fully managed platform designed for the era of AI. 

Key takeaways:  

Admins can confidently govern agent development at scale using Security Hub and PPAC, with tools like environment rules, authentication enforcement, and granular sharing controls. 

Makers can build responsibly by using Security Hub insights and Microsoft Information Protection (MIP) labels to ensure their agents protect sensitive data across Dataverse and Microsoft Copilot Studio. 

Admins can use Security Hub’s network and connector management capabilities, like IP firewall, Azure vNet, and connector policies, to prevent data exfiltration and securely scale AI across their enterprise. 

With layered network controls, virtual network support, and tenant isolation defaults, organizations can confidently adopt AI while ensuring customer data privacy and strict boundary protection. 

With agents now powering mission-critical workflows, Managed Availability ensures your Copilot Studio agents stay online with seamless failover, zone redundancy, and enterprise-grade resiliency, protecting against regional outages and infrastructure failures. 

Admins can confidently manage and recover agent-driven solutions using Self-Service Disaster Recovery (SSDR), safe deployment practices, and proactive monitoring, ensuring agents remain responsive, integrated, and business-critical even during disruptions. 

Stay Tuned: More to Come Shortly

We will have more in-depth product demos over the next two weeks covering topics including: Agent cost controls, Databricks knowledge, Dataverse MCP, Code Interpreter, BYOM, new Knowledge sources for agents, and more. Be sure to find them on Power Platform YouTube Channel. 

Links to all the sessions mentioned above:  

GenAI for Enterprise: Intelligent Apps and Agents with Dataverse & MCP | BRK129 

Knowledge in Copilot Studio | OD812 

Unlock your Dataverse data with secure and powerful search | DEM523 – YouTube 

Agents with real-time knowledge from Databricks – YouTube 

Tools for agents | OD811 

Agents in action: Document processing 2.0 | OD814 

Enhancing Agent Management Controls for AI-Ready Enterprises | BRK132 

Dataverse for agents | OD813 

Enhancing Agent Management Controls for AI-Ready Enterprises | BRK132 

Building safe and reliable Microsoft Copilot Studio agents | OD817 

To learn more and get started, check out these useful links:  

Watch the opening Build keynote: Microsoft Build 2025 | Satya Nadella Opening Keynote 

Read the Build blog: Announcing new Microsoft Dataverse capabilities for multi-agent operations | Microsoft Copilot Blog 

Learn about Dataverse: What is Microsoft Dataverse? – Power Apps | Microsoft Learn 

Listen to the Podcast: Low Code Approach | Microsoft 

Read about Copilot Studio: Customize Copilot and Create Agents | Microsoft Copilot Studio 

See what’s new in Copilot Studio: What’s new in Copilot Studio | BRK140 

The post Dataverse at Build: The Agent Platform Powering the Future of Agentic AI appeared first on Microsoft Power Platform Blog.
Source: Microsoft Power Platform