New Microsoft database, analytics tools target agentic AI

New capabilities

Agents are the current vanguard in enterprise AI.

Microsoft, like fellow tech giants AWS and Google Cloud and data platform vendors including Databricks and Snowflake, has made AI development a priority since OpenAI's November 2022 launch of ChatGPT, which represented a significant improvement in generative AI (GenAI) capabilities.

Many enterprises boosted their investments in AI development after ChatGPT was released because GenAI can make workers better informed and more efficient. With data the foundation for AI, many data management and analytics vendors responded by creating environments and added capabilities aimed at better enabling customers to build AI applications.

After assistants and copilots initially were the most advanced GenAI applications, agents have emerged over the past year to become the major trend in enterprise AI.

Microsoft's latest database and analytics capabilities -- in conjunction with numerous other new features unveiled during Build -- aim to better enable developers as AI evolves, according to Frank Shaw, Microsoft's chief communications officer, who spoke during a virtual media briefing ahead of the conference.

"Our goal is to help developers turn their ideas into reality," he said.

Azure Cosmos DB is a NoSQL database in Fabric, Microsoft's AI-powered data management and analytics suite.

Using Azure Cosmos DB, which includes an integration with the Azure AI Foundry AI development environment, users will be able to bring semistructured data such as JSON files and certain documents into Fabric alongside their structured data to provide a more comprehensive foundation for analytics and AI-driven analysis.

"Cosmos DB in Fabric is a big addition to bring enterprise-grade NoSQL capabilities into Fabric, enabling organizations to incorporate semistructured operational data alongside their structured and [providing] developers flexibility to build modern AI applications … with just a few clicks," Catanzano said.

Chat with your data, meanwhile, is a new feature in Power BI, Microsoft's main analytics platform, which enables users to discover and deploy Fabric agents to aid data discovery and analysis. The feature expands natural language processing (NLP) capabilities beyond questions about a single report while it's open on a user's screen to queries across multiple reports, semantic models, applications and agents.

Given that it significantly broadens the AI-powered chat capabilities of Power BI, it's a valuable addition for users, according to Michael Ni, an analyst at Constellation Research.

"By embedding agents into Power BI, Microsoft isn't just adding another conversational BI feature -- it's activating a decision layer," he said.

Beyond Azure Cosmos DB and chat with your data, Microsoft introduced the following database and analytics capabilities, all of which are in preview:

• Enriching custom-built agents developed in Copilot Studio with prebuilt Fabric data agents that can use data from OneLake -- a data lake in Fabric -- to discover insights, trigger actions and automate processes within channels such as Teams and 365 Copilot.
• The use of Azure AI Foundry to create agents built on data stored in Azure Databricks.
• Digital twin builder in Fabric to make it easier and more efficient to develop digital twins that can be used for applications such as scenario planning.
• PostgreSQL database capabilities in GitHub Copilot to accelerate the development of GenAI applications.
• SQL Server 2025, including easy integrations with Azure and Fabric to streamline development.
• Shortcut transformations in OneLake to automatically transform data to the Delta Lake table format or apply AI-powered data transformations, such as document classification, as data gets ingested to prepare it for analysis or development.
• Translytical capabilities, a feature in Power BI that enables users to automate actions from within reports.

The intent of the new database and analytics capabilities is to provide Microsoft users with a broad array of tools from which they can choose as they develop AI applications, according to Kevin Scott, Microsoft's chief technology officer.

"We're going to give you a ton of tools to work with, and we hope you do something with them," he said during a virtual media briefing.

But beyond simply providing users with an array of new analytics and database tools, Microsoft is delivering them in an integrated manner that makes sense, according to Ni. They're not piecemeal. They're designed to work together -- particularly within Fabric -- and augment one another.

"Microsoft's latest announcements shift Fabric from a data and analytics platform to a runtime for decisions, AI agents and intelligent action," he said. "They're removing the friction between data, insight and action, providing developers with the infrastructure to build intelligent, agentic apps from day one."

Regarding the highlight feature beyond Azure Cosmos DB in Fabric and the chat capabilities in Power BI, enabling developers to combine custom-built agents with Fabric agents and integrate them throughout the Microsoft ecosystem is noteworthy, Ni continued.

"Combined with Cosmos DB inside Fabric, Microsoft is fusing real-time operational context, expanded Fabric capabilities to orchestrate data from across the enterprise, and Power BI semantics to deliver low-friction conversations with all the data a user has access to -- at scale and in context," he said.

Looking ahead

While Microsoft's latest database and analytics capabilities provide developers with new capabilities that foster agentic AI development, Catanzano noted that they do not directly address the large amounts of unstructured data that can make AI models and applications more accurate.

Unstructured data such as text, images and audio files now make up more than 80% of the new data being created. While structured and semistructured data can provide enterprises with insights into their operations, the addition of unstructured data gives a more complete portrayal. In addition, with AI applications requiring large amounts of data to draw upon to reduce the occurrence of hallucinations, unstructured data can be valuable.

Vendors such as Databricks and Snowflake are among those also making access to unstructured data priority in AI development.

"Most other vendors are highly focused on dealing with the mountain of unstructured data," Catanzano said. "Some of this covers it, but others are more broadly talking about their capabilities around it."

Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than 25 years of experience. He covers analytics and data management.