Last month I watched a founder demo a product built almost entirely with AI. Onboarding worked. Dashboard looked polished. Pitch deck matched the UI. Investors nodded.
Then someone asked: "What happens when a new user has no data yet?"
Silence. No empty state. Error messages were generic. Pricing promised features that weren't in the build. It looked finished. It wasn't.
A lot of founders are living in that moment right now.
The illusion of done
AI lowered the floor for what "shipping" means. Screens, copy, code, even a brand identity in an afternoon. The speed is real. Speed without direction creates debt that doesn't show up until a customer hits it.
Three pressures at once:
- Investors and users expect more, faster. "Looks legit" got harder because everyone else's MVP looks legit too.
- Teams are smaller. You're not hiring a full product org at seed, but you're competing with products that feel full-sized.
- AI fills gaps unevenly. Excellent at producing something. Weak at deciding what should exist, what should wait, and what will confuse someone on day one.
Result: products that launch quickly and stall on retention, conversion, or trust. Often the idea wasn't bad. Nobody owned the experience end to end.
I use AI daily in Cursor when I'm building. It drafts. I still have to decide whether Hermeneia's Simple vs Deep Insight modes make sense for a confused first-time reader, or whether Trova's waitlist copy overpromises safety features that aren't ready.
What a product designer does that AI doesn't
I'm not arguing AI is useless. The difference is what you point it at.
A product designer is not "the person who makes things pretty." They're the person who turns ambiguity into decisions the team can build against:
- Frame the problem before the solution. What job is the user hiring this for? What are they afraid of? What do they need to believe before they click pay?
- Design for states, not just screens. Empty, loading, error, success, permissions, mobile vs desktop: where products actually break.
- Prioritize. Not everything on the roadmap deserves a sprint. Cut scope without cutting value.
- Keep product, marketing, and engineering aligned. One visual language. One story. One set of behavior rules.
AI can draft options for all of that. It can't tell you which option matches your user, your timing, or your business model. That judgment is the product.
When to hire or contract
You don't need ten designers. You need design leadership early enough that you're not rebuilding after your first hundred users.
Bring someone in when:
- You're about to raise or launch publicly and the product still feels like a prototype
- Engineering is blocked because requirements keep changing mid-sprint
- You're getting signups but not activation: people arrive and don't know what to do next
- You've shipped three versions of the same feature because nobody agreed on "done"
A fractional or contract product designer can cover a lot: audit the experience, define a focused roadmap, build a lightweight system engineers can reuse, and stay close through launch.
What you can do this week
- Write the user story for the worst case. Confused, impatient, skeptical. Not the happy path.
- Record five people using the product. Don't explain. Watch where they hesitate.
- Separate demo quality from product quality. A screen that wins a meeting is not always a screen that wins a customer.
- Give someone ownership of the experience. If it's split across founder, engineer, and marketer with no single owner, it drifts.
AI made it possible for everyone to build. It did not make it easy to build the right thing. Treat design as strategy, not paint at the end, and you ship less noise.
The landing page is easy. The product is where you win. That's the work I do.