Microsoft Unleashes ‘Hey Copilot’ & Autonomous Agents Across All Windows 11 PCs | Anthropic Boosts Enterprise AI with ‘Skills’ & Competing Agent Commerce Protocols Emerge
Key Takeaways
- Microsoft rolls out voice-activated ‘Hey Copilot’ and experimental autonomous ‘Copilot Actions’ to all Windows 11 PCs, aiming to redefine the operating system experience.
- Anthropic introduces ‘Skills’ for Claude, allowing enterprises to create reusable, specialized AI expertise packages, significantly boosting workflow efficiency and consistency.
- The future of AI commerce faces a critical juncture as Google, OpenAI/Stripe, and Visa unveil competing agent payment protocols, raising concerns about interoperability and trust.
- Strella secures $14M to scale its AI platform, accelerating customer research by conducting AI-moderated voice interviews that yield faster, more candid insights.
Main Developments
Today marks a pivotal moment in the evolution of artificial intelligence as Microsoft announced a sweeping transformation of Windows 11, integrating voice-activated AI assistants and experimental autonomous software agents across every PC running the operating system. With “Hey Copilot,” users can now summon Microsoft’s AI by voice from anywhere in Windows, positioning voice as the third fundamental input method after the mouse and keyboard. Complementing this, Copilot Vision expands globally, allowing the AI to analyze on-screen content and provide contextual assistance, now with full document context in Office applications. Most ambitiously, Copilot Actions, an experimental feature for Windows Insiders, will enable AI to take control of a user’s computer to complete tasks autonomously, operating within a sandboxed environment with rigorous security measures like “agent accounts” and explicit user opt-in. This move signifies Microsoft’s aggressive bet on embedded, agentic AI to redefine personal computing, moving beyond chatbots to AI that takes action.
This push towards agentic AI is not limited to the desktop. In the enterprise, Anthropic is empowering Claude with a new capability called ‘Skills.’ This feature allows users to create reusable folders of instructions, code scripts, and reference materials that Claude can autonomously load when relevant, effectively acting as an “evolving playbook” for specialized expertise. Unlike traditional prompt engineering or RAG, Skills employs a “progressive disclosure” architecture, enabling Claude to intelligently navigate and utilize vast amounts of information without overwhelming its context window, resulting in significant efficiency gains—Rakuten, for instance, reported an 8x productivity improvement in finance workflows. This approach provides a consistent, customizable, and composable method for enterprises to integrate deep domain knowledge into their AI agents.
The rapid advancement of agentic AI, however, brings its own set of challenges, particularly in securing transactions. The dream of AI agents buying things for people is currently hampered by the lack of a universal trust infrastructure. This week saw the emergence of three competing agentic commerce protocols: Google’s Agent Pay Protocol (AP2), OpenAI and Stripe’s Agentic Commerce Protocol (ACP), and Visa’s Trusted Agent Protocol (TAP). Each aims to establish a trust layer that convinces banks, merchants, and customers alike that AI agents can securely handle finances. While these protocols are crucial for enabling AI-driven commerce, industry observers express concern over the potential for “walled gardens,” where different chat platforms and their associated payment protocols may not interoperate, creating confusion and fragmentation for enterprises.
Amidst these transformative shifts, the underlying robustness of AI agents remains paramount. A new framework from Stanford University and SambaNova, called Agentic Context Engineering (ACE), directly addresses the critical challenge of “context collapse” in LLMs. ACE treats an agent’s context window as an “evolving playbook,” automatically populating and modifying it with structured, incremental updates rather than compressing information. This prevents the degradation of an AI’s memory over time, ensuring agents remain comprehensive, efficient, and consistent. Proven to outperform existing methods, ACE provides a foundational technology for building the very self-improving, robust AI agents that Microsoft is deploying, Anthropic is enhancing, and future commerce protocols will rely upon.
Further demonstrating the immediate impact of AI agents, Strella, an AI-powered customer research platform, raised $14 million after just one year in stealth. Strella’s technology compresses eight-week research projects into days by using AI to moderate voice-based interviews. Crucially, studies show participants are often more honest with AI moderators than with humans, leading to richer insights. With mobile app screen sharing and robust fraud detection, Strella is not only making customer research faster and deeper for clients like Amazon and Chobani but also democratizing access to customer feedback across entire organizations, enabling new research practices that wouldn’t have been possible before.
Analyst’s View
Today’s news underscores an undeniable acceleration into the era of agentic AI, where systems move beyond conversation to autonomous action. Microsoft’s integration of Copilot’s voice and agent capabilities directly into Windows 11 represents a massive leap toward mainstream adoption, but simultaneously elevates the stakes for security, privacy, and user control. The simultaneous emergence of competing agent commerce protocols highlights the industry’s struggle to build a secure and interoperable transactional layer—a critical hurdle that could lead to fragmentation rather than seamless experiences. The race to develop robust, trustworthy agents, as exemplified by Anthropic’s ‘Skills’ and Stanford’s ACE framework, will be key to unlocking their full potential. The coming months will test the industry’s ability to govern these powerful new capabilities and establish common standards, determining whether AI agents deliver on their promise of unprecedented productivity or become entangled in a new kind of platform war.
Source Material
- ACE prevents context collapse with ‘evolving playbooks’ for self-improving AI agents (VentureBeat AI)
- Amazon and Chobani adopt Strella’s AI interviews for customer research as fast-growing startup raises $14M (VentureBeat AI)
- Microsoft launches ‘Hey Copilot’ voice assistant and autonomous agents for all Windows 11 PCs (VentureBeat AI)
- How Anthropic’s ‘Skills’ make Claude faster, cheaper, and more consistent for business workflows (VentureBeat AI)
- Google vs. OpenAI vs. Visa: competing agent protocols threaten the future of AI commerce (VentureBeat AI)