As the enterprise software industry rapidly reshapes around AI agents, competition among related companies to lead the early race also appears to be intensifying. In the process, an AI data war has emerged as a hot issue.

More companies are blocking outside firms' products from accessing their platforms. Microsoft, a big business software company, has joined the trend.

According to a recent report by The Information, Databricks began testing a new feature in early March to help customers connect information on its platform more easily to visualisation tools, but it was blocked by Microsoft. Databricks customers were unable to access Microsoft's Power BI product, which is used by almost all Fortune 500 companies.

The feature Databricks sought to test focused on enabling Power BI customers to manage data and easily build AI agents on Databricks rather than using Fabric, Microsoft's data management solution. With Fabric excluded in that scenario, the situation could be uncomfortable for Microsoft.

Microsoft says it blocked data management tools, including Databricks, from connecting to Power BI for reliability reasons. But outside observers see it as part of a fight over control of semantic layer tools, which have emerged as a strategic point for running AI agents more accurately and cheaply.

AI agents need clear, well-structured data to function properly. Semantic layers that help generate context-rich data can play a key role in developing AI agents that accurately handle multi-step tasks, The Information reported.

Some business software companies blocking rival agents from accessing customer data in their products, or seeking to allow access only for a fee, could also be a move aimed at semantic layers.

Microsoft competes with Snowflake, Databricks and others in the data management software market. Shehab Amin (셰합 아민), CEO of startup LakeSail, which helps AI agents process data quickly, told The Information: "The fact that it cut off the Databricks connector alone makes it clear where Microsoft sees the battleground for the next-generation platform war." He added: "For the past 30 years, Microsoft's strategy has been to capture the product where customers start their work, then use that as a base to lead them to buy other Microsoft products as well."

AI agents are seen as a new growth engine in the enterprise software industry, but a view persists that AI agents could threaten these companies' core businesses.

A scenario in which customers use outside companies' AI to analyse internal data, rather than the software vendor that holds that data, is something enterprise software companies would not want to happen. An analysis says Microsoft's share price falling 25 percent from last year's peak is not unrelated to such a scenario.

Microsoft is also very aggressive on AI agents. To reduce the need for customers to use outside companies' AI agents, it added AI agent functions to its flagship Microsoft 365 platform and introduced Work IQ, which supports companies in pulling data from Microsoft applications to use in other apps. It has also recently launched the Agent 365 platform that allows companies to centrally manage AI agent environments. Agent 365 supports not only Microsoft but also third-party AI agents.

Against this backdrop, Microsoft's rivals are speeding up efforts to support companies in using semantic layers without being tied to a particular software vendor. Snowflake and Salesforce are leading Open Semantic Interchange, a cooperative network involving nearly 50 companies to develop an industry standard for semantic layers. Open Semantic Interchange includes Amazon Web Services and Oracle, but not Microsoft, The Information reported.

SAP, a large business applications company, is also accelerating an AI agent-focused strategy under the banner of an 'Autonomous Enterprise', but it is passive about allowing outside agents to access data on its core platforms.

SAP has focused on routing external AI agents through its AI assistant Joule rather than allowing direct access to its platform. In section 2.2.2 of its API policy updated in April 2026, SAP explicitly banned AI systems from independently scheduling or executing API calls. That means external AI agents cannot use SAP APIs. At its annual Sapphire 2026 conference, SAP Chief Customer Officer Thomas Saueressig (토마스 자우어에식) defended the API policy as governance needed for a multi-tenant platform.