The news: Anthropic has launched a research preview that allows its Claude chatbot to directly control a user’s Mac through Claude Code and Cowork, marking a significant escalation in the race to build AI agents capable of executing tasks independently, per Venture Beat.

The system can click buttons, open applications, type into fields, and navigate software on behalf of the user, effectively transforming Claude from a conversational assistant toward a remote digital co-worker role. 

For teams, the pricing starts at $20 per seat per month for groups of five to 75 users. Enterprise pricing is custom and includes administrative controls to toggle Cowork on or off for the organization.

Why it’s worth watching: The release thrusts Anthropic into the center of a fiercely competitive market that includes OpenAI, Google, Microsoft, and a growing swarm of startups, all racing to deliver agents that work inside existing tools rather than assisting alongside them. 

Tests show capabilities and constraints: 

• Early hands-on testing suggests Anthropic’s solution succeeds about 50% of the time, per Venture Beat, handling information retrieval and summarization reliably but struggling with complex multiapplication workflows. 
• That current success rate and unresolved audit trail gaps suggest organizations should limit deployment to low-risk, highly repetitive tasks where failure carries minimal consequence. 

Zooming out: US employees are already using AI for a range of workplace tasks. More than 4 in 10 use AI to edit work (42%), prepare presentations (42%), and draft emails to clients (41%). Significant shares also rely on AI to write internal emails (35%), create strategy (34%), and outline work (34%).

These are the kinds of routine but time-sensitive tasks that Anthropic’s agent could theoretically execute without human prompting, though it is not yet reliable enough to perform them autonomously. 

Implications for brands: As agentic tools become mainstream, successful adoption will hinge on governance: defining which workflows are appropriate for agents and enforcing those boundaries.

As AI evolves from assisting to acting, control architecture may become a competitive differentiator. Organizations that pair automation with robust oversight will lead; those that don’t will inherit unmanaged risk.