Feb 17, 2026
Turning a 6 week design project into 2 hours with Claude Cowork
Product Designer at DOSS
The interaction I had between Claude Cowork and Claude Code yesterday was nuts. Has me kicking myself for not really diving deep into AI tooling ~6-12 months ago.
I've been working on a bunch of prototype explorations for the "V2" evolution of the DOSS UX and UI. One of the things that has been really tricky about this is that it's really hard to make a reasonable prototype exploration without a proper imitation of our actual data model, and the complexity of our customers' implementations.
I'd been trying to do this on my own, and then I hit a wall. My understanding of the nuances of those data structures just isn't sophisticated enough (yet).
Setting the Stage
To build a realistic prototype, I needed a hypothetical customer that could give me a good range of scenarios to design against. I chose Thursday Boots.
They're similar to one of our most complex real customers, Verve Coffee, and they gave me the range I was looking for. They're mostly online D2C, but just opened their first brick and mortar.

The day before I launched the workflow below, I described Thursday's business to Claude and had it scrape the entire Thursday Boots website.

From that, it put together a complete database of all their SKUs, a range of personas in the Thursday business that would likely touch our product in their daily work, and what parts of our software would be most relevant to them — including how they'd likely use them in their real work settings.
That gave me the foundation. The next day, I went deeper.
Going Deep with Cowork
I pointed Cowork at the repository of all of the documentation, system design specs, implementation details, customer calls, etc. for one of our most recent, most complex customer installs — Verve Coffee.
Here's what I told Claude:
"Go read over all of this documentation. Watch all of these Fireflies recordings & read through the transcripts. Spin up as many sub-agents as you need to in order to develop a thorough understanding of this entire corpus of material. Then produce a doc with a detailed synthesis of your research.
Once you're done with that, go back over all of our discussions about the direction for this V2 structure, and all of the discussions we've had about this hypothetical customer we've created — the implementation spec, the customer personas, the daily workflows for each persona, and all of the ways those would show up in the system.
You're going to synthesize both of these together. Help me identify the gaps in my understanding.
Then, extrapolate everything you've learned about this real DOSS customer install into a series of changes to the data model for this prototype. Highlight all of the specifics of how our data model is going to change based on the structure of a live DOSS customer. Make this another MD doc that we can reference.
Then put together a plan and to-do list for Claude Code.
Create a new branch in our repo, take all of the output of the initial branch and put it into an archive doc so we don't pollute the context of this next phase."
The Result
Had to make 2 minor edits to the docs Cowork produced.
Then I handed the plan and to-do list to Claude Code. Executed flawlessly.
Used to be that this would have been a ~6-week project. Took maybe 2 hours.





