The prompt writes the screen. It won't tell you if the screen should exist.

Speed is the pitch you keep hearing. Ship in an afternoon. Prototype while you sip your coffee. Cut the design team to a tenth of its size. It sells well on a slide and it makes any CFO happy.

But a year of real work says the pitch misses the point. Let me catch you up on where the value actually lives, and what your team still owns that no prompt touches.

Faster gets you the same product, sooner

Dan Maccarone spent a year rebuilding his studio's design process around AI, across four real products with real clients. His finding, stated flat: it didn't make us faster. Sprints took five days before. They still take five days.

What you get when you optimize for raw speed is the same product, just sloppier and shortsighted. He calls it high-speed mediocrity, work that shows up fast because nobody slowed down to think. The bots can generate a screen in ninety seconds. They can't tell you whether that screen should exist.

That gap is your job. The person using the product never sees your sprint. They see whether it does the thing they showed up to do.

Where the tools actually earn their keep

The real gift is depth, not speed. Maccarone's team was rebuilding a subscription media product with five user states on every page, across desktop, tablet, and mobile. The old way was a relay: design desktop, review, approve, then mobile, then tablet, then the edge cases. A single product ran for a month.

This time they built all five states at once. Live and interactive. Flip a toggle and the page rearranges for a logged-out stranger versus a power subscriber. Same five days, an order of magnitude more product. The depth that used to blow out schedules now fits inside the sprint.

The documentation shift matters just as much. Instead of a stale PRD, a Figma file, and a Slack thread that all drift apart, the docs and component library generate from the approved prototype. Change a flow and the acceptance criteria regenerate to match. The working build becomes the single source of truth everything else answers to.

Your loudest skeptic is telling you something true

When Maccarone pitched a three-day sprint, his COO of twenty years told him straight it wasn't going to happen. He was right. That skepticism is the most useful thing in the room, and founders keep treating it like an obstacle to steamroll.

Worker trust in company AI tools actually fell as the tools got more capable, not the other way around. You don't fix that with a mandate or a pep talk. You keep the skeptic close, hand him the job of poking holes in every step, and let him watch the output get better. Doubt turns into quality control. One day he stops arguing with the process and starts defending it.

Craft still starts with a person, not a mood board

The same lesson shows up outside software. When Love & War had to brand a food hall inside the 110-year-old Lord & Taylor building on Fifth Avenue, the key move came before any color or typeface. They named it Shaver Hall after Dorothy Shaver, the store's forgotten former president. A real name arrives with built-in texture and a reason for people to care. She did the heavy lifting decades before the agency showed up.

That is problem framing, and it starts with research, not design. The same instinct runs through the craft world. Hideki Yoshimoto built Craft x Tech only after standing in a lacquerware workshop and watching a soup bowl made like a sculpture. His curator, Maria Cristina Didero, puts it plainly: design is about people, not about chairs.

Sanne Visser makes it a system, redesigning the process behind materials rather than a single product. The judgment about what to make, and why, sits with the humans.

The deep cut

The trap is that the docs now generate from the prototype and stay in sync. That sounds like a win, and it is. But it also means the prototype is the decision. If your team builds a screen nobody framed, the system will faithfully generate user stories and acceptance criteria for a screen that shouldn't exist, and it will all look finished and correct.

So move your framing to the front and guard it. The brief gets written and approved by engineering before anyone opens an AI tool. The five days you save on relay work go straight into deciding what not to build. That decision is the one thing the tool will never make for you, and now it's the only thing that matters.

Three questions for your team

1. On our next project, who owns the "should this exist" call, and does that framing happen before anyone opens an AI tool?

2. Who is our loudest skeptic, and have we given them a real validation role instead of talking past them?

3. When we redesign a legacy product or brand, are we starting with research to find the person or story hiding in it, or reaching for a mood board first?