Build 2026: Organizations Can Unlock Enterprise Intelligence with Microsoft IQ

Microsoft has made Microsoft IQ generally available across GitHub Copilot, Microsoft Foundry, and Copilot Studio. Alongside the announcement, the company announced a new family of in-house AI models and platform enhancements aimed at making enterprise AI smarter, more secure, and easier to manage at scale.

Microsoft IQ is a unified intelligence layer that connects AI, enterprise data, and organizational context, and it gives AI agents a deep understanding of how a business actually operates. It consists of four main components: Work IQ, Work IQ APIs, Fabric IQ, and Foundry IQ.

Work IQ learns from things like emails, meetings, chats, and documents to understand how tasks are performed and what users need. This allows AI tools to give smarter suggestions and even anticipate actions. Work IQ APIs allow developers to build custom AI solutions that can automate tasks or design intelligent assistants tailored to specific roles or workflows.

On the other hand, Fabric IQ organizes business data into clear, meaningful structures so AI can interpret it accurately instead of treating it as raw, disconnected information. Foundry IQ handles knowledge (such as company documents, policies, and stored content) and ensures AI retrieves the right information securely and in context. The Work IQ APIs will become generally available for commercial customers on June 16.

Web IQ is a new addition to the Microsoft IQ family that improves how AI agents search and use information from the Internet. It acts as a fast, AI-focused web search layer that helps agents quickly find the most relevant pieces of information from real-world sources.

New Microsoft in-house models expand reasoning, vision, and speech capabilities

Microsoft announced today several new in-house models, including MAI‑Thinking‑1. This first
reasoning-focused model is built entirely on high-quality, licensed enterprise data to handle complex, multi-step tasks, long-context processing, and coding efficiently. It is currently available in limited preview through Microsoft Foundry.

Additionally, Microsoft has released several specialized models for different tasks. MAI‑Image‑2.5 supports both generating images from text and modifying existing visuals, which makes it useful for creative workflows. Other additions include MAI‑Transcribe‑1.5 for accurate multilingual speech recognition, MAI‑Voice‑2 for advanced voice outputs in multiple languages, and MAI‑Code‑1, which is an efficient coding model integrated into tools like GitHub Copilot and VS Code.

Fireworks AI integration strengthens Microsoft Foundry ecosystem

Last but not least, Microsoft is expanding its AI ecosystem by bringing Fireworks AI into Foundry. It enables developers to work with different models on a single platform while still meeting enterprise requirements like governance and data residency.

Additionally, Frontier Tuning uses reinforcement learning within secure boundaries, which allows AI agents to improve over time using a company’s own data, workflows, and knowledge. Agent 365 for local agents extends Microsoft Entra, Purview, and Defender into a single control plane, allowing IT admins to observe, govern, and secure agents across their estate.