Microsoft Web IQ Brings Bing Search Data to Agent Applications

Microsoft has expanded its IQ suite of products with the introduction of Web IQ, an agent-focused web search tool that leverages the massive Bing search index to provide up-to-date general information for use in applications.

Web IQ builds on the company's existing IQ suite, which uses data as a set of embedding vectors and integrates it with Model Context Protocol (MCP) servers to tie LLM output to business data, reducing the risk of hallucination and improving accuracy.

By treating web search data as a sparse vector store needed for grounding data in LLM-based applications, Web IQ provides access to fresh information beyond what's used during model training. This is particularly useful for agents operating in ecommerce or weather-sensitive product management, where related information from the wider world can be linked to business data.

The platform uses the DiskANN algorithm developed by Microsoft Research, allowing fast search without requiring enormous amounts of in-memory data access. Web IQ structures the retrieved information as 'structured evidence objects' and passage-level information from unstructured text documents, delivering a higher signal-to-noise ratio than querying a search engine.