Microsoft used its Build 2026 conference this week to emphasize a pivotal shift: AI agents are increasingly being integrated into enterprise systems, with the most successful platforms poised to be those offering robust context, governance, identity management, memory, and secure data access.
Key announcements included Microsoft IQ, designed as a contextual layer across GitHub Copilot, Microsoft Foundry, and Copilot Studio. Additionally, the company revealed Work IQ APIs slated for release on June 16, Fabric IQ for structured business data, Foundry IQ for knowledge retrieval across enterprise data and the live web, and Web IQ, a novel web search stack tailored for AI agents.
Further expanding its AI capabilities, Microsoft introduced Scout, a personal work agent, alongside seven new in-house AI models within its MAI family, encompassing various modalities and use cases, including the advanced MAI-Thinking-1.
These developments are squarely in the domain of Marco Casalaina, Microsoft’s VP of Products for Core AI and an AI Futurist. He leads Microsoft’s AI Futures team and has previously overseen teams for Azure AI, including Azure OpenAI, Vision, Speech, Decision, Language, Responsible AI, and AI Studio. Before joining Microsoft in early 2022, Casalaina led Salesforce’s Einstein AI team and holds a computer science degree from Cornell University.
Understanding Casalaina’s Role and the “AI Futurist” Title
Casalaina explained his dual role: “I am VP Products of what we call Core AI. Core AI is our set of tools for AI developers, and that includes Foundry, Visual Studio, VS Code, GitHub and GitHub Copilot.” His title as an “AI Futurist,” he clarified, has a practical implication: “I am the first person to try anything new here. I am constantly getting things from all over Microsoft… I get to see a lot of really cool stuff.” He frames his futurist perspective as focusing on the immediate future, approximately one year ahead.
Microsoft’s Strategy for Agentic AI Adoption
Addressing the surge in enterprise adoption of agentic AI, Casalaina outlined Microsoft’s multi-layered approach: “At the very base of the stack is our commitment to model choice.” This includes established models from OpenAI and Anthropic (with Claude Opus 4.8 now available on Foundry), alongside Microsoft’s own burgeoning MAI models, developed for “token efficiency, optimization and customization” for clients’ proprietary data sets.
Moving up the stack, Microsoft is introducing “hosted agents in Foundry” for managed agent capabilities, handling scaling and containerization. The Foundry control plane offers crucial oversight for these agents, monitoring costs, token usage, and performance through continuous evaluation.
The centerpiece of the Build announcements are the “IQs,” designed to provide specific contextual intelligence to agents:
- Foundry IQ: For accessing and processing unstructured knowledge.
- Fabric IQ: Enables agents to interact with data stored in Microsoft’s cloud services like Fabric and Power BI, bypassing traditional reporting interfaces.
- Work IQ: Acts as an agentic interface for Microsoft’s ecosystem, including Outlook, Teams, Word, and SharePoint.
- Web IQ: A new, high-speed, headless web search capability optimized for agent interactions, capable of searching the web, videos, and performing browsing tasks.
Complementing these is “Agent Optimizer,” featuring new evaluation methods for assessing agent performance and enabling modifications to improve accuracy, creating a vital feedback loop.
The Purpose and Integration of “IQs”
Casalaina clarified that the “IQs are headless,” primarily serving developers and the agents they build by providing distinct types of context. While end-users may not directly interact with the IQs, their functionality is becoming visible, as seen with Microsoft 365 Copilot leveraging Work IQ.
He confirmed that “All of the IQs are indeed exposed as MCP servers,” characterizing MCP as an “agent-facing or self-describing API.” This architecture is crucial for secure access, particularly for Work IQ, which needs authenticated access to user data within the Microsoft ecosystem. This ties into Microsoft’s announcement of agent identity management through Entra, allowing agents to possess their own identities, inboxes, and communication channels.
Microsoft’s Multifaceted AI Strategy: Models, Infrastructure, and Connectivity
Addressing Microsoft’s position in the AI landscape, Casalaina stated, “The answer is yes” to being a model company, infrastructure provider, and connector. Microsoft remains committed to offering a wide array of leading models from partners like OpenAI, Anthropic, and Mistral, while concurrently developing its proprietary MAI models. These in-house models are optimized for “token efficiency, bang for the buck and customization,” addressing key customer demands for fine-tuning and continued pre-training capabilities.
He emphasized the importance of data provenance for their models, assuring customers of clean data rights. Furthermore, Foundry supports various agent orchestration frameworks, such as LangGraph, LangChain, and CrewAI, integrating them as first-class Foundry agents with full observability and evaluation capabilities.
Future of Model Usage and Agentic Workflows
Casalaina anticipates an expansion in the number of models enterprises utilize, driven by the nuanced value derived from each token and the overall cost-effectiveness. He highlighted substantial growth in user adoption, citing over 20 million monthly active users for Microsoft 365 Copilot alone, and enterprise deployments like Bayer’s agent system used by 20,000 employees.
Illustrating practical applications, he shared examples of agents managing grid operations for the Australian Energy Market Operator (AEMO) and enabling human empowerment by reducing overwhelming alert volumes. The focus is shifting from simple question answering to AI meaningfully assisting with complex work.
Observability, Governance, and ROI in Microsoft Foundry
The Foundry control plane is central to these aspects, offering detailed insights into agent token usage, costs (which can vary based on model routing), and integration with Azure’s comprehensive cost management. Governance is further enhanced through new rubric-based evaluation systems, allowing for granular testing of agent functionality beyond basic grounding, ensuring agents perform specific tasks correctly and reliably.
The Crucial Role of Copilot and User Engagement
Microsoft 365 Copilot is presented as a significant advantage, providing a familiar interface for users to access custom-built agents. The ability to “publish to Copilot” seamlessly integrates specialized agents into users’ existing workflows within Teams and other Microsoft applications, significantly boosting adoption and utility.
Casalaina detailed how Copilot, powered by Work IQ, has transformed his own productivity. He recounted using it to draft a complex customer email by referencing past conversations and documentation, saving him approximately an hour of manual effort. This reflects the broader shift toward AI taking actions and handling time-consuming information retrieval tasks.
AI’s Impact on Personal and Professional Productivity
Casalaina expressed confidence in AI’s positive impact on his own role, stating, “My job has changed. For one thing, I do a lot more now, both in my business life and personal life.” He cited using Web IQ for a personal task, like locating a specific car model across a wide geographic area, saving him hours of research. This demonstrates AI’s potential to augment both professional and personal efficiency.
A Call to Developers
Casalaina concluded with a clear message for developers: “Foundry is really the place. This is the place where you can build your agents, scale your agents, test your agents and improve your agents. That’s what it’s all about, and it’s happening.”
Business Style Takeaway: Microsoft’s strategic push towards integrated AI agents with its ‘IQs’ and Foundry platform signifies a maturation of enterprise AI, moving beyond standalone models to deeply embedded tools that enhance productivity and data accessibility. This underscores the critical need for businesses to invest in platforms that provide context, control, and seamless integration, positioning AI agents as essential operational components rather than novelties.
Based on materials from : venturebeat.com