Claude Cowork for legal teams

• Mark, an in-house product lawyer at Anthropic, demonstrates how Claude Cowork helps him quickly come up to speed and respond to a product manager's question without rereading all his old work.
• He relies on a scheduled morning task that acts like a chief of staff, plus a custom /brief skill (built on Anthropic's legal department playbook) that pulls the relevant review, reads new Slack and Gmail threads, and surfaces just what changed.
• The workflow keeps a human in the loop — Mark verifies the cited paragraphs, approves Claude's drafted reply, and closes the Jira ticket — letting him spend time on strategic thinking instead of busy work.

• Claude Cowork can run a scheduled task first thing in the morning that delivers a personalized memo about what's on your plate, what's new, and what's urgent, with a connected Gmail inbox sorted by what's due today.
• The /brief skill is what Mark runs to come up to speed on a specific product fast; it already knows where the review file lives and how their templates are structured, so the prompt can be short.
• The skill is built as an open-protocol plugin based on how Anthropic's own legal department works, and anyone can open it up and tailor it to their own company's playbooks.
• Instead of reading a 40-page memo to find three paragraphs, the brief points Mark at exactly the relevant paragraphs: what was concluded before, what's changing, and which parts of the analysis it touches.
• "Trust but verify" is core to the job — Mark clicks into the earlier review paragraph the brief quoted to confirm it's there verbatim, because his name goes on the reply.
• Keeping a human in the loop means Claude drafts the reply to the product team and asks Mark to approve it before sending, preserving human judgment.
• Tracking work in Jira lets Mark close out the ticket so the team has context next time, building a shared corpus of legal knowledge instead of information silos.

Claude Cowork — A Claude product that helps professionals schedule tasks, run skills, and connect tools like Gmail and Jira to handle their work. Skill — A reusable capability (here /brief) that encodes a workflow and knows where files live and how templates are structured. /brief — The skill Mark runs to quickly come up to speed on a specific product and change. Plugin — A packaged, open-protocol extension (the brief skill) that can be customized to a company's own playbooks. Scheduled task — A task Claude Cowork runs automatically, e.g. a morning memo of what's due and urgent. Human in the loop — Keeping a person to review and approve outputs (verifying citations, approving the drafted reply). Trust but verify — The practice of checking Claude's cited source material because your name is on the final reply. Corpus of knowledge — A shared, accessible body of legal context built up over time instead of siloed information.

Hi, I'm Mark. I'm an in-house product lawyer at Anthropic.

This morning, a product manager sent me a Slack asking a quick question on a feature that launched a few months ago. Now, I've only got a few minutes for my next meeting and none of the context I had when I wrote the original memo about it. The old me would have spent the first hour of the day rereading my own work, pulling up the old brief, figuring out what actually changed, and then start thinking about the question. Here's how I set it up in Claude Cowork so I don't have to do that anymore.

With Claude Cowork, I can schedule a task to run first thing in the morning. It's like having a personalized chief of staff that delivers a memo for me each day about what's on my plate, what's new, and what's urgent. And my Gmail inbox is connected to this one. I've got five items sorted by what's due today. Two are routine, one's new, and one's that follow-up on the product that the product manager requested. That's what I'm clicking on.

The skill I lean on most is /brief, which is what I run when I need to come up to speed on a specific product fast. I helped build this plugin based on how Anthropic's own legal department works. It's open protocol. Any of you can open it up and custom tailor it so it's based on your own company's playbooks. I type /brief and name the product and specific change. The skill already knows where our review file lives and how we structure our templates. So my prompt is short, just what I'm actually trying to figure out. Claude pulls the review from the folder, it reads the new Slack and Gmail threads, and gives me back what I need in the shape that I think in: what we concluded back then, what's actually changing, and which parts of the whole analysis it touches. I'm not reading a 40-page memo to find three paragraphs. The brief points me at those three paragraphs.

I click into the earlier review paragraph the brief quoted, and there it is, verbatim. I'm checking because my name goes on the reply, and trust but verify is pretty much the whole job. We're ensuring that we keep a human in the loop. Instead of doing busy work or work about the work, spending time pulling up all these documents and searching for files, I can use my brain to do more strategic thinking. That's why I went to law school in the first place.

Now, once I feel comfortable with my guidance, I ask Claude to draft a reply to the product team, and Claude asks me to approve it. We track tickets in Jira at Anthropic, so I can also have it close out our ticket. That way the rest of my team has context the next time a question comes up about this topic or this product area. And we're building out a corpus of knowledge so that anybody in the legal department, or if needed, the rest of the company, can access that information instead of building information silos. It's really empowering.

Case closed. I can walk into my next meeting feeling confident that I was responsive to a fast-moving product team. Right now, a lot of lawyers are being asked to do more with less and embrace AI. With Claude Cowork, I feel like I'm able to do exactly that.