Anthropic released computer use for Mac as a research preview this week, making Claude the first frontier AI assistant with native macOS desktop control available to paying subscribers.

The feature lets Claude open apps, click buttons, type text, navigate the file system, and complete multi-step tasks on a user’s Mac without manual oversight. It is available now to Pro and Max subscribers on Anthropic’s platform, with broader access expected to follow.

How It Actually Works

Claude’s computer use is not a screen-recording bot or macro system. It works through a layered decision chain. When given a task, Claude first checks whether there is a direct integration available, a purpose-built connector for Gmail, Slack, Notion, or similar tools. If one exists, it uses that. If not, it falls back to browser automation. Only as a last resort does it invoke screen control to interact with arbitrary desktop apps.

This architecture matters because it means Claude defaults to structured, auditable paths and only uses visual control when nothing else is available. That is a meaningful safety design choice: the system prefers traceable integrations over ad-hoc clicking.

The feature was built in part using technology from Vercept AI, a startup Anthropic acquired in February 2026. It is paired with a companion feature called Dispatch, which lets users assign tasks to Claude from their iPhone and have them execute on a connected Mac.

What This Changes

For individual professionals, the implication is straightforward: tasks that currently require your attention can now be delegated to an AI that sees the same screen you do. Research, data extraction from desktop apps, reformatting documents, moving files, filling forms: all of it becomes delegatable.

For businesses, the implication is broader. Most enterprise software was designed assuming a human would click through it. Computer use means that assumption no longer holds. An AI agent can interact with any application, not just the ones that have been integrated via API. That dramatically expands what is automatable, particularly for businesses running older internal tools, complex ERP workflows, or multi-application processes that were never built to be connected.

The timing matters too. Microsoft released Copilot Cowork (a similar agentic desktop assistant built in part on Anthropic’s Claude) the same week. Google has had computer use capabilities in Gemini Enterprise since late 2025. The market is converging on the same product bet: AI that does work in your existing environment rather than requiring you to move into a new one.

The Agent Workforce Picture

This is a different kind of AI capability than chat or document generation. Those tools augment a person doing work. Computer use replaces the person doing the work, at least for well-defined tasks.

For the businesses Enterprise DNA works with, this is the capability most likely to change day-to-day operations over the next six to twelve months. Not because every task will be automated, but because the class of tasks that can now be delegated to an AI agent has expanded significantly. Anything a human can do by looking at a screen and clicking through a workflow is now within range.

The practical question is not whether this technology works. The question is whether your business has identified which workflows are worth targeting first, and whether you have someone guiding that prioritisation before your competitors do.

What This Means for Business

A few direct implications worth tracking:

Older internal tools become automatable. Businesses running legacy software that lacks an API are no longer locked out of AI automation. If a human can operate it by looking at a screen, so can Claude.

The definition of “AI integration” has changed. Previously, connecting AI to a business system required developer work to build an API integration. Computer use means AI can interact with systems that have no integration at all. This lowers the barrier to AI adoption significantly.

The pace of change is compressing timelines. Claude computer use went from zero to production-ready in the same month as competing features from Microsoft and Google. Businesses that are still evaluating whether AI is ready are watching these capabilities ship while they deliberate.

The research preview designation signals this is not Anthropic’s final word on the feature. Refinements, enterprise-specific controls, and expanded access are likely over the coming months. The companies that begin testing it now will understand its limits and possibilities well before those who wait for general availability.

---

If this is the kind of problem agents can help with, the free Working With Claude field guide is the practical next step. Thirty-two pages, no fluff. Get the free guide.