A day after "/design" shipped,
Samsung had aproduction result.
Claude Code's "/design" command, a terminal-based UI mockup generator, shipped on August 19. The very next day, Samsung reported it had cut a month-long screen-design process down to two days using Claude Code. Unlike past headline cases such as Uber and Salesforce, mostly from outside Korea, this one is unusual for landing right after the feature shipped.
A case study, arriving
with unusual speed
Claude Code added a "/design" command on August 19, letting developers generate UI mockups directly in the terminal. Screen design had long meant handing work off to an external tool like Figma, with a constant back-and-forth cost between engineers and designers exchanging files.
One day later, Samsung Electronics' mobile software organization announced that adopting Claude Code for One UI's screen-design process had cut what used to take about a month down to two days. According to Anthropic's customer story page, the scope covered was the workflow from wireframing new One UI screens through engineering review. The Korea Herald also covered the case the same day.
A day after a feature ships,
a real-world result from amajor manufacturer lands.
From scattered cases abroad
to a major electronics maker
Past enterprise case studies leaned on companies like Uber and Salesforce, mostly outside Korea.
Until now, the enterprise adoption stories showcased for Claude Code were mostly software companies abroad, like Uber and Salesforce. Samsung's case is an example of the same effect showing up inside an internal engineering org at a company that isn't a software business first. It's also plausible that the bigger the organization — and the more entrenched the division of labor between engineers and designers — the larger the payoff from cutting that back-and-forth.
There's also something to the sheer speed of the case study's publication, just one day after the feature shipped. Anthropic appears to be building a pipeline for collecting and publishing customers' production results right after a feature launches — a sign that the competition with OpenAI and GitHub Copilot is playing out not just on feature-release speed, but on how fast impact can be demonstrated.
Who this matters to,
and how
Engineers / dev teams
This shows the back-and-forth with designers can plausibly be resolved inside the terminal. A practical first step is trying "/design" on one small screen.
Leadership / investment decisions
A concrete "~1 month → 2 days" figure is usable in an internal budget request for AI tooling — but be precise about the scope it covers (wireframe through review only).
PMs mid-pilot
An early, large-scale adoption case from another company is useful leverage for proposing an internal pilot expansion.
If you're considering this,
scope it down first
Testing small and measuring the effect before a company-wide rollout is the standard playbook here too.
Try one screen type
Rather than rolling out across every product at once, try "/design" on one recurring screen type — a settings screen or list view — and measure the time difference against your current flow.
Bring designers into the review
Don't evaluate with engineers alone — structure the pilot so designers can review it too, checking visual quality and brand-guideline compliance.
Decide where the saved time goes
Deciding upfront which downstream step (extra rounds of user testing, for instance) absorbs the freed-up time makes the case stronger when you take it to leadership.
"Two days" covers only
part of the process
Don't let the number run away from you. The "two days" in Anthropic's case study covers only wireframing through engineering review — it doesn't include user testing, final visual-design polish, or accessibility checks. Reading it as "the whole screen-design process now takes two days" would be wrong.
Company-published case studies also skew toward successes by nature. The same effect isn't guaranteed to reproduce in a different team or company. If you're actually evaluating adoption, the practical move is running a small pilot against your own workflow first, verifying the effect, and only then expanding.