From Chatbots to an AI Computing Layer
Microsoft AI agents are software entities that can understand context, make decisions, and execute multi-step tasks across apps and devices with minimal human prompting, turning AI from a question-answering chatbot into an active computing layer that coordinates work, information, and services in the background. At Microsoft Build 2026, Satya Nadella made it clear this is the company’s next big bet. Instead of AI being an add-on inside Office or Windows, Microsoft now talks about agents that sit above and across operating systems, files, and hardware. This shift reframes AI as the primary way users interact with digital systems, similar to how graphical interfaces replaced command lines. The focus on Autopilots, MAI models, and new hardware signals that Microsoft wants AI agents to orchestrate workflows, not just respond to prompts, and to run wherever people are working.
Autopilots and Scout: AI as a Proactive Enterprise Colleague
Autopilots enterprise automation is Microsoft’s new label for customizable AI agents that behave more like junior employees than passive assistants. The flagship example is Microsoft Scout, an “always on” agent built on OpenClaw that works across Teams, Outlook, OneDrive, SharePoint, and the wider Microsoft 365 stack. Instead of waiting for commands, Scout can prepare meeting briefs, juggle calendars, and track projects in the background. Microsoft is also stressing identity and security: each agent receives its own Entra identity so IT teams can define what the agent can see and which actions it may take. According to Techloy, this marks a shift from chatbots that answer questions to systems “designed to proactively handle workplace tasks.” With agents able to execute multi-step workflows inside OS-enforced boundaries, Microsoft is positioning Autopilots as the next layer of enterprise automation, not a novelty alongside existing tools.
MAI-Thinking-1: Reasoning Power for Agentic Workloads
If Autopilots are the visible layer of Microsoft AI agents, MAI-Thinking-1 is the reasoning engine meant to drive them. MAI-Thinking-1 is a 35-billion-parameter model with a 128,000-token context window, designed for complex multi-step instructions, long-context reasoning, and code generation. Microsoft highlighted its lower token cost compared to similar models, signaling an intent to make large-scale reasoning more affordable for everyday workflows. The model anchors a broader MAI family, including MAI-Image-2.5, MAI-Transcribe-1.5, MAI-Voice-2, and MAI-Code-1, all available through Microsoft Foundry and core apps like PowerPoint and OneDrive. This is not only about better answers; it is about agents that can remember and reason across long projects, documents, and codebases. As Techloy notes, Microsoft is reducing its dependence on OpenAI and building an in-house AI stack that spans models, infrastructure, and hardware.
Project Soltera and Solara: Agents That Follow You Across Devices
Nadella’s vision for an AI computing layer extends beyond Windows. Project Soltera (also described as Solara) is a chip-to-cloud, Android-based agentic OS designed for “an open, multiple agent world.” Microsoft displayed two concept devices: a Qualcomm-powered wearable badge that keeps agents accessible away from the laptop, and a desk companion built to “think, plan, and get things done without breaking flow.” Techloy describes Solara as a platform where agents move across devices, tasks, and environments so computing feels less like opening apps and more like having continuous assistance. This approach puts Microsoft in direct competition with phone, PC, and wearable platforms that want to own the next interface. If Soltera-style devices succeed, AI agents will no longer live inside a single browser tab or chat window but become ever-present, coordinating personal and enterprise workflows throughout the day.
RTX Spark and Quantum: Hardware for an Agent-First Future
To make this AI computing layer real, Microsoft is tying agents to dedicated hardware. The Surface Ultra and the compact Surface RTX Spark Dev Box are both designed as AI-first machines built around Nvidia RTX Spark chips. Microsoft says the Dev Box can run AI models of up to 120 billion parameters locally, easing reliance on cloud GPUs and giving developers a fast way to prototype and deploy Microsoft AI agents on their desks. At Build, Nadella appeared with Nvidia’s Jensen Huang to underline this partnership, even as he stepped away from heavy dependence on OpenAI. Looking further ahead, the Majorana 2 quantum chip and the Microsoft Discovery research platform hint at a future where quantum and classical AI infrastructure blend. Together, these moves show Microsoft treating AI not as a feature inside Windows, but as foundational infrastructure spanning silicon, operating systems, and agents.
