Azure AI Search Introduces Agentic Retrieval for Enhanced Relevance

Microsoft's Azure AI Search has rolled out agentic retrieval, an advanced query engine that autonomously manages retrieval strategies for complex questions, enhancing answer relevance by up to 40% in conversational AI. This innovation uses conversation history and Azure OpenAI to decompose queries into subqueries executed parallelly across text and vector embeddings, accessible via the new Knowledge Agents object in the 2025-05-01-preview REST API and Azure SDK prerelease packages. The system integrates a dedicated "Agent" resource linked to Azure OpenAI, employing both keyword and semantic search capabilities, with results reranked into a unified payload. The public preview, available in select regions, offers free initial per-token billing for Azure OpenAI's query planning and Azure AI Search's semantic ranking. Microsoft envisions a future where AI agents from different companies collaborate effectively, supporting the Model Context Protocol (MCP) for creating an "agentic web" and emphasizing improved memory for AI agents to enhance user interactions. Kevin Scott, Microsoft's CTO, highlighted the significance of structured retrieval augmentation and secure interactions through platforms like SSOJet, which offers API-first solutions for authentication processes.