I designed and shipped a real product this year. Baraxis — a business management platform for mobile bar operators. Brand, product design, marketing site, front-end, backend. One designer, an AI-augmented workflow, and a stack I'd have needed a small team for two years ago.
It's live. People pay for it. So I can talk about what AI actually changed, instead of guessing.
There's a hype version of this story and a doom version. The hype version says AI means anyone can build anything now. The doom version says AI is producing a flood of broken software nobody can maintain. Both are wrong in the same way: they assume the tool does the thinking. It doesn't. It does the typing.
What it actually changed
The honest answer: AI collapsed the distance between knowing what to build and having it built.
I've always been able to design a product. What I couldn't do, fast, was ship the whole thing — the auth, the database, the payment flow, the PDF generation, the email notifications, the hundred unglamorous pieces between a Figma file and a working product. That gap is where most designers' product ideas die. AI closed it.
Claude Code wrote most of the implementation. I'd describe a feature — invoice builder, send a branded PDF, notify on payment — and it would scaffold it. Cursor handled the tighter edits. Supabase for the backend, Stripe for payments, Next.js for the front-end. The work that would have meant hiring a developer, or abandoning the idea, became something I could do in an evening.
Where it would have sunk me
Here's the part the hype skips. AI is confidently, fluently wrong on a regular basis. It will write code that runs and is also subtly broken. It will solve the wrong problem beautifully. It will make an architectural decision that works for ten users and collapses at a thousand.
The only reason the AI output was usable is that I could tell when it wasn't. I have an IT degree. I know what a sane database schema looks like, what a security hole smells like, when a quick fix is going to become a permanent problem. Every time the AI handed me something, I was the one deciding whether to ship it, edit it, or throw it out and start over.
That's the whole job now. Not typing the code. Knowing which code deserves to exist.
A junior designer with the same tools would have shipped something that looked done and fell apart in month three, because they wouldn't have known what to reject. The tool didn't make me senior. Being senior is what made the tool useful.
What this means for the work
I think we're watching the designer-engineer boundary dissolve, and I don't think it's coming back.
For years, design and engineering were separated by a wall of implementation cost. A designer had an idea, an engineer built it, and the handoff between them was where ambition went to get value-engineered down. That wall was expensive. AI just made it cheap. The designers who can think across the whole thing — strategy, brand, product, code — aren't doing three jobs anymore. They're doing one job that used to need three people.
That's not a threat to senior designers. It's the best thing that's ever happened to us. The ceiling on what one person can ship just went through the roof. The only thing standing between an idea and a shipped product is whether you know what good is and can recognize it when the machine hands it to you.
I knew what good was. So I shipped.