AI Pulse · AI Trends Pulse

Anthropic just rolled out “Record a Skill” inside Claude Cowork, and it’s worth paying attention to if you’ve ever wished your team could bottle up how they actually do things. You screen-record yourself working through a task, talk out loud as you go, and Claude watches. It then converts that demonstration into a skill it can run again, no further prompting needed. The feature is live now for Cowork users.

This matters because most business knowledge lives in people’s heads, not in manuals. The finance person who knows exactly which cells to check before closing the books, the ops manager who can spot a dodgy supplier invoice in three seconds, the marketer who preps campaign reports the same careful way every week. Recording a skill means you show the AI once, and it can repeat the process when you need it. No scripting, no hoping someone wrote it down correctly.

Why this fits the command centre idea

We’ve been building exactly this pattern into the Omni Command Centre for months. The principle is the same: capture tacit process knowledge from the people who do the work, turn it into something repeatable, and let an agent handle the routine execution. Anthropic’s approach is screen-and-voice. Ours layers in structured workflows, data connectors, and guardrails so the skill doesn’t just run, it runs inside your actual systems with the right permissions and audit trail.

The broader point is that AI is moving from “answer my question” to “do my recurring task”. If you’re still thinking of AI as a chatbot, you’re already behind. The companies that will pull ahead in the next year are the ones turning their best operators’ instincts into skills an agent can execute reliably. That’s not futurism. It’s shipping now.