Claude Cowork GTM Workshop
There were 900+ people on the waitlist for this event. That number tells you something. When WorkOS and Anthropic put together a hands-on workshop for Claude Cowork — their agentic desktop tool for GTM teams — the SF tech community showed up. Hard.
I got in. And I'm glad I did.
The Room Was the Right Room
These are the events I live for — not because of the catering or the venue (though the WorkOS crew does not miss), but because of who ends up in the room. Founders, engineers, operators, and the kind of people who book calls with strangers they just met and actually follow through. I walked out with a handful of new connections I've been genuinely nerding out with since — shoutout to Martijn Lancee, Mark Robinson, Igor Shvartser, and Will Reese. This is what the tech community does at its best.
What Actually Happened in the Workshop
Zack Proser from WorkOS ran the workshop, and he has a gift for making complex things click. The curriculum was GTM-focused: prospect research, personalized outreach at scale, and maintaining a live spreadsheet that tracks the whole loop. Very sales-y on the surface — not exactly my daily domain as a UX designer.
But then the spreadsheet materialized from thin air.
One moment: nothing. Next moment: a living, structured artifact, built by an agent that had just done the research, drafted the content, and organized everything into a shareable doc. I've understood agentic workflows intellectually for a while. Watching it happen in real time, in a room full of people audibly reacting to the same thing — that's different. That lands differently.
I need to be honest: this workshop didn't change how I think about agentic workflows. It confirmed that the way I've been thinking about them is correct. The autonomy runs in the background. The human stays in control. The interface is the trust layer. There it was, live, in a GTM context.
One Thing I Noticed as a Designer
Claude Cowork has the workflow built right in — prominently, intentionally. I use multiple AI tools and the UX philosophy varies a lot across them. What struck me here was the intentionality of the design: the agent loop isn't buried in settings or tucked behind advanced mode. It's the product. The interface is built around the assumption that you are going to run multi-step autonomous tasks — and it scaffolds that for you.
That's a design decision. And it's the right one.
The Part Where I Went a Little Feral
Here's the thing about watching a workflow automate itself: your brain immediately starts mapping it onto everything else.
I spent the drive home mentally rebuilding the same loop for UX work. Research a product space → synthesize patterns → draft positioning or test scripts → track everything in a live artifact. By the next morning I had a rough workflow sketched out. By the end of the week I was annoying my own group chats with "okay but what if we automated THIS part too."
It's a little bit of a drug. I'm not going to pretend otherwise.
Lydia's Q&A
The evening closed with a Q&A with Lydia Hallie, Member of Technical Staff at Anthropic, and she teased a few things coming down the road for Claude that I'm genuinely keeping an eye on. Can't say more than that because she didn't say more than that — but the direction is exciting and I'm ready to experiment the second it ships.
One Footnote
Meeting the WorkOS team in person made me want to build something just to have an excuse to use their product more. I have an experimental project brewing. Stay tuned.
If you missed the workshop, the recording and slide deck are available — WorkOS posted them after the event. Highly recommend watching Zack's demo section even if you skim the rest.
