The modern professional day is rarely defined by the high-level strategic work listed in a job description. Instead, it is consumed by a relentless tide of administrative friction: the status update that needs to be synthesized from three different Slack channels, the onboarding checklist that must be updated for a new hire, and the endless formatting of slide decks. This is the invisible tax of knowledge work, a layer of operational noise that separates a specialist from their actual expertise. For years, the promise of AI was to automate the core task, but the reality of the corporate workflow suggests that the real bottleneck is not the thinking, but the coordination.

The Shift to Ambient Agency

Anthropic is betting that the future of AI is not a chat box you visit, but a coworker that follows you. The company has officially expanded Claude Cowork from a desktop-only experience to a full mobile and web ecosystem. Currently rolling out as a beta for Max subscribers, this expansion transforms the tool from a destination into a persistent layer of the professional environment. The technical architecture is designed for continuity; a user can initiate a complex multi-step project on a laptop and transition to a mobile device to monitor progress. Crucially, the system now supports background processing, meaning the agent continues to execute tasks even after the user closes the application, effectively decoupling the AI's productivity from the user's active screen time.

This rollout comes with a revealing set of data that challenges the prevailing narrative of the AI agent market. After analyzing 1.2 million sessions across more than 600,000 organizations, Anthropic discovered that the primary driver of Cowork's utility is not technical development. While the industry has largely viewed agents as tools for automating software engineering, the data shows that the vast majority of Cowork users are not coding. The largest share of activity falls under business processes and operations, accounting for 33.4% of usage. These users are primarily employing the agent to consolidate fragmented updates into cohesive reports or to construct operational checklists.

Content creation and copywriting follow as the second most popular use case at 16.4%. In stark contrast, the software development tasks that dominate the AI discourse represent only 8.7% of the total volume. Other technical categories trail further behind, with DevOps and infrastructure at 7%, research and intelligence at 6.4%, and data analysis and business intelligence at 5.8%. When aggregated, nearly half of all Cowork activity is dedicated to general office administration and operational management, signaling a massive, underserved market of non-technical knowledge workers.

The Strategy of the Peripheral Task

Anthropic describes this phenomenon as the work around the work. This is the critical distinction between a professional's core competency and the connective tissue required to move a project forward. A lawyer's core competency is legal analysis, but the work around the work is the formatting of filings and the coordination of discovery documents. By positioning Claude Cowork as a tool for these peripheral tasks, Anthropic is avoiding the friction of trying to replace the expert, instead focusing on removing the administrative burdens that distract the expert from their primary value.

This approach is part of a calculated two-track product strategy. On one end, Anthropic offers Claude Code, a terminal-based tool designed for the deep, focused work of building and debugging software. On the other end, Claude Cowork targets the broader, more diverse population of global knowledge workers. This strategic divergence appears to be paying off in the enterprise sector. According to the Ramp AI Index released in May, Anthropic's enterprise adoption rate hit 34.4%, overtaking OpenAI's 32.3% for the first time. The data suggests that while OpenAI may lead in general consumer awareness, Anthropic is successfully penetrating the corporate structure by solving the specific pain points of organizational operations.

Under the hood, this capability is powered by the Claude Sonnet 5 model. The improvements in reasoning and tool-use capabilities in Sonnet 5 allow Cowork to handle complex, multi-stage workflows that require a high degree of autonomy. However, to address the primary fear of enterprise leaders—the loss of control—Anthropic has implemented a human-in-the-loop architecture. Through mobile notifications, the agent can autonomously handle the bulk of a task but will pause to request human approval at critical decision points. This transforms the user experience from a synchronous chat, where the human prompts and waits, to an asynchronous collaboration, where the AI prepares a draft in the background and the human simply reviews and approves it.

For organizations integrating these tools, the lesson is that the point of adoption has shifted. The goal is no longer just to find a tool for the developer or the marketer, but to identify the common connective tasks that exist across every single department. The transition to a web-based version further lowers the barrier to entry, particularly in corporate environments with strict software installation policies. The real efficiency gain comes from distinguishing between core expertise and peripheral noise. When an organization can isolate the work around the work and delegate it to an agent, it doesn't just increase speed; it restores the professional's ability to actually perform the job they were hired for.

The era of the chatbot is ending, replaced by an era of asynchronous agents that treat the corporate calendar as their primary interface.