Microsoft adds data agents to Fabric to help business users with analytics

Microsoft is adding new data agents to Fabric, its cloud-based offering combining data warehousing, data science, data engineering, and data analytics services.

Its goal is to help business users find insights in enterprise data without the help of a business analyst.

Most enterprises either don’t have business analysts or their analysts are too busy to deal with all the demands for their services, said Arun Ulag, corporate vice president of Azure Data at Microsoft.

That’s why self-service analytics tools like those Microsoft is adding to Fabric are necessary, he said.

The new data agents will be available across all Fabric services, including Synapse and PowerBI. They can be created from within any service and configured by providing them with access to the data the user wants to analyze.

The agents can access data held in Fabric’s OneLake storage, or data elsewhere using over 200 connectors, Ulag said.

Natural-language interaction

Triggered by natural-language queries, the agents can complete tasks such as retrieving data, reasoning, and creating data visualizations. Ulag said the agents should be able to deal with prompts such as, “Can you analyze all my customer survey questions, comments, focus on all the negative comments, categorize them into the top five buckets and give me examples?”

That kind of natural-language interaction is key, said Arnal Dayaratna, research vice president at IDC.

“The innovation here is that these data agents can provide conversational assistance to any enterprise knowledge worker that is interested in performing data-driven analytics,” he said. “This enables a radical democratization of the capability not only of data analysis, but also the skill of explaining the significance of data-centric insights.”

While such agents could be relevant for any enterprises, Forrester principal analyst Noel Yuhana sees them delivering the most significant value to large enterprises grappling with complex, multi-platform data environments.

Nucleus Research’s Duncan Van Kouteren further narrowed down the list of likely adopters to enterprises with existing Microsoft ecosystem investments, noting that the company is following an industry trend of targeting such agents at business analysts and decision-makers with less technical expertise.

Other comparable offerings include Salesforce’s Einstein Copilot, Google Cloud’s agents built via Vertex AI Agent Builder, IBM’s Watsonx Assistant, agents inside Snowflake Cortex, and a combination of OpenAI GPTs, custom tools and RAG pipelines.

Data agents can be customized via Copilot Studio

The data agents built inside Fabric automatically show up inside Microsoft’s Copilot Studio and can be used as base templates to build more customizable agents that can complete more tasks without manual intervention, Ulag said.

The agents’ interactive nature means “They can correspond with your business users; They can update systems as well among other things,” Ulag said.

Van Kouteren foresees that the Copilot Studio integration could enable domain-specific agent customization.

Dion Hinchcliffe, lead of the CIO practice at The Futurum Group, sees applications for agents customized in Copilot Studio across supply chain, finance, and HR operations.

“In supply chain operations, a custom Copilot could proactively monitor logistics data in OneLake and trigger alerts if shipping delays are likely. In Finance operations, enterprises could automate forecasting conversations with an agent grounded in real-time data and ask questions of it like, ‘What will Q3 expenses look like if hiring increases by 10%?’”

Data agents are currently in public preview. Part of the Fabric subscription, they will use up Fabric credits when interacted with.

Risk of lock-in

Hinchcliffe said the data agents follow the classic “land and expand” strategy for Microsoft.

“In the short-term, these agents will drive higher Fabric and Azure AI consumption (and value delivery), bundling value into existing E5 or Fabric SKUs to accelerate usage,” Hinchcliffe said.

“In the mid-term, Microsoft will look to monetize high-value, vertical-specific copilots via Copilot Studio customizations and the Azure AI Agent Service,” he added.

In the longer run, he sees Microsoft trying to position Fabric as the control plane for enterprise data agents, making it indispensable to enterprise AI orchestration.

“This will support higher Azure usage, seat-based Copilot licensing, and enterprise lock-in,” he said.