Azure Updates: Storage Actions; AI Foundry; Fine-tuning models

Microsoft VP of Azure Storage Aung Oo announced that the company is making Storage Actions generally available. The managed platform is aimed at transforming the way customers manage tasks with Azure Blobs and Data Lake Storage. Currently, many customers are struggling with data sprawl and a lack of uniform standards for data management, attempting to manage data as best as possible with disparate tools. This often involves creating custom scripts for smaller datasets, using automation runbooks, or developing with different SDKs. Errors along the way can result in data loss. 

According to Oo, with Storage Actions, customers can rapidly compose a task with complex conditions to target objects based on their properties. By taking this approach, customers can reduce complexity, enhance efficiency, and improve consistency.

VP of Product for AI Foundry Yina Arenas shared insights on multi-agent apps with the open Agent2Agent protocol. The new protocol is coming to AI Foundry and Copilot Studio to support collaboration across clouds and platforms. “We believe that Microsoft Copilot will empower every employee and act as the ‘UI for AI’ to connect with agents and agentic systems—networks of agents that reason, act, and adapt across boundaries. As customers scale these systems, interoperability is no longer optional. They want their agents to orchestrate tasks that span vendors, clouds, and data silos. They want control, visibility, and trust—without being locked in,” Arenas wrote.

A2A will support structured agent communications to exchange goals, manage states, invoke actions, and provide results. For developers, Microsoft emphasized the ability to use well known tools like LangChain and Semantic Kernel. Each call is routed through multiple safety systems including Microsoft Entra, AI Content Safety, mutual TLS, and audit logs.