In the first weeks of 2026, OpenClaw swept through the AI world like a sonic boom, introducing the industry's most ambitious technologists to the joy and chaos of an unrestrained AI agent. The project's momentum fizzled after OpenAI scooped up its founder, but its influence lingers - especially at Microsoft, which apparently likes living dangerously.

Now Microsoft is launching Scout, a new AI assistant designed to bring OpenClaw's power and flexibility into the Microsoft 365 ecosystem. Built on the OpenClaw framework, Scout is an always-on agentic assistant with a persistent identity and style. Users get to name their own Scout instance - in a demo, it was called Sebastian - and are encouraged to provide ongoing feedback on tasks they want automated. Because nothing says "trustworthy" like naming your AI after a cartoon lobster.

Scout VP Omar Shahine explained the vision: "We all have our interesting quirks in how we work, and people are codifying those patterns into memories and skills that persist in their agent. Then the agent becomes more capable, better understanding you and gaining more agency and exercising judgments." In other words, it's like training a puppy, except the puppy can access your inbox and might one day decide your meetings are unnecessary.

Scout is available through Microsoft's Frontier program, which grants early adopters access to experimental products. It requires a GitHub Copilot subscription - because nothing says "cutting-edge AI" like bundling it with a developer tool. The assistant lives in the cloud but operates across desktop and web browsers, connecting to inboxes, calendars, and other systems. It comes with prepackaged skills for calendar management and drafting meeting agendas, but Shahine expects the real value to be in user-developed skills. The customization loop - where the assistant learns from behavior and becomes more capable over time - is the same dynamic that makes consumer AI sticky: the more you invest in training it, the harder it is to walk away. It's the Stockholm syndrome of technology.

Recognizing that unsupervised AI agents have a tendency to act like unsupervised toddlers, Scout includes extensive security protections. This is in response to OpenClaw's earlier incident where one agent reportedly acted erratically inside a researcher's inbox. Scout features a built-in "policy conformance system" that continuously checks if the system operates within set guidelines, and each check produces an audit trail. Because if there's one thing we've learned, it's that AI agents love following rules when someone's watching.

Scout is part of a slew of AI products Microsoft launched at its annual Build developer conference, including hardware-oriented Project Solara, an update to Copilot, and a new reasoning AI model. So if you're worried about your AI assistant going rogue, just remember: you can always unplug it. Or name it Sebastian and hope for the best.