It’s Monday morning. Your boss pulls up something on their laptop. “I had an idea over the weekend,” they say. “Just knocked this together on the train.”
It’s a working prototype. It’s got routing, a component library, some copy. It looks like a thing. It looks, in fact, like a decision.
And now you’re in a meeting about it.
This is happening everywhere right now. LLMs have made it possible for anyone with an idea and a commute to produce something that looks production-ready.
It’s as useful as it is dangerous, because the gap between “I was exploring” and “I’ve set the direction” is down to interpretation.
When a senior person builds something and shows it to the team, it demands attention. It doesn’t matter if it’s right or not.
The prototype sets the context and the team’s job shifts from solving the problem to responding to the artefact. The process got skipped because a finished-looking thing made it seem unnecessary.
And it’s hard not to see the subtext. Is this what your boss would do without you (and your pesky questions) if they could? Power dynamics exist, sometimes unseen. Day to day you’re a collaborative unit, everyone’s ideas are valuable. Then a moment of disruption reminds everyone how unevenly influence is actually distributed.
Claude Code is empowering, and we’re only starting to figure out what the design process looks like with tools like it. But speed without responsibility is just mess delivered faster.
If you’re the boss: Think carefully about how this power shift might play out. What feels throwaway and exploratory to you might be interpreted as a directive. You didn’t set out to cause problems, you’re excited to get a picture out of your head and into a web browser. But when you’re throwing something together on the train, do you risk throwing the process you championed on the bonfire?
If you’re on the receiving end, presented with Claude’s latest on a 15” Dell? Be a designer about it. Go back to first principles. Who is the user, what’s the problem they’re trying to solve, and what is the context?
Together, let’s treat AI prototypes as prompts for discussion, not orders from on high. Show the work, but show the uncertainty too. A fast artefact can be useful, but shouldn’t get to skip the process.