I Rebuilt My Whole Interface Without a Designer or a Coder

How two AI tools — one that draws and one that builds — let a one-person company redesign an entire product in six weeks, and what that means if you run a small business.

By Mogens Rye, Founder of MyAvatar · June 2026

Six weeks ago, the part of my platform that customers actually see and click — the screens, the buttons, the dashboards — looked like a product built by one person on a budget, because it was. Today it looks like a product built by a design studio and a front-end team. Nothing in between involved hiring either.

I want to walk through how that happened, honestly and in plain terms, because it is one of the clearest examples I have of where AI has genuinely changed what a small company can do. Not a forecast. Something I did this quarter.

What redesigning an interface used to cost

Start with the old reality, because it is the thing that just broke.

If you wanted to redesign a software product properly, you needed two scarce, expensive skills. First, a designer — someone who decides what each screen looks like, how it is laid out, which colours and typefaces carry the brand. Then a front-end developer — the person who turns that picture into working code that runs in a browser. (“Front-end” simply means the part of the software you see and interact with, as opposed to the machinery humming away behind it.)

For a small company, that is a wall. A designer and a front-end developer in Copenhagen are months of salary, or an agency invoice that makes you wince. So most small products never get redesigned. They limp along looking like what the founder could manage, while better-funded competitors look polished. I lived on the wrong side of that wall for a long time.

Two tools, two jobs

What changed is that I now work with two AI tools that, between them, cover both of those scarce skills.

The first is Claude Design. It is a tool with a chat box on one side and a blank canvas on the other. I describe the screen I want — “a post-login home page built around a few large action tiles, calm, with our brand colours” — and it draws a real, working mock-up on the canvas. I can then refine it by talking to it, or by pointing at a specific element and adjusting it, until the screen looks right. It is, in effect, the designer.

The second is Claude Code. It is an AI that reads and writes the actual code my platform runs on. Once a design looks right, Claude Design hands it over to Claude Code with essentially one button — the finished design passes straight across — and Claude Code builds it into the live product. It is, in effect, the front-end developer.

The design tool draws it. The coding tool builds it. I decide whether it is right.

That last part is the bit that matters, and I will come back to it. But the shape of the workflow is the headline: a screen goes from a sentence I type, to a picture I can judge, to working code in the product, without either of the two specialists I could never afford.

I am not a designer, and I am not a front-end developer. I am a founder with a clear idea of what good looks like — and, it turns out, that is now enough.

What we actually rebuilt

This is not a hypothetical. Let me show you the real trail, because the volume is the point.

It started in the middle of May with the Co-pilot — the home page you land on after logging in. We rebuilt it from scratch around a set of large action tiles: instead of dropping you into a wall of features, it asks, in effect, what do you want to do today? and routes you to the right place. Each of those routes — make a video, distribute content, find leads, run a campaign, check analytics — became its own guided front door.

That single redesign did something quietly important: it gave birth to a proper design system. A design system is best thought of as one named master-list of the product’s colours, fonts, spacing, and components. Change the list, and every screen that uses it updates in step. Ours has a settled character now — a fresh lime-green on warm, paper-like backgrounds, serif headlines with clean sans-serif labels, flat and calm, no fussy shadows. Before, every page reinvented its own look. Now there is a shared language, and new screens speak it automatically.

From that foundation, the rest came fast. Over the following weeks we built and shipped, among others:

• A family of specialist workspaces, each fronted by a named AI agent with its own personality: Magnus on the sales floor, Nadia running distribution, Carl directing the movie studio, Hugo on content, Iris on analytics, and Ava acting as a project director who turns a goal into a step-by-step plan. These share one underlying kit, so they feel like a team that belongs together rather than nine unrelated tools.
• A proper customer-record system — clean account and contact overviews and a deal pipeline.
• A Voice Studio, with onboarding wizards that walk a new user through setting up their own voice and video.
• A public price calculator — you name your budget and it shows the best video that fits — and a guided Google Ads campaign builder.
• And, literally today, a Show Studio for creating video-podcast episodes.

Every one of those is a real, live surface. Several came through a formal “handoff” — Claude Design produced the mock-up, packaged it, and passed it to Claude Code to build. The trail of those handoffs is still sitting in the codebase, which is how I can tell you this with confidence rather than from memory.

The discipline that made it work

It would be dishonest to make this sound effortless, so here is the part people skip.

Speed like this only stays useful if it stays consistent and safe. Two disciplines made that true.

The first is the shared design system I mentioned. Because every new screen draws from the same master-list of colours and components, ten screens built across six weeks still look like one product, not ten. Without that, fast would have meant chaotic.

The second is the review loop. Claude Code does the building, but it builds on a test version of the platform first, not the live one customers use. I look at the result, decide whether it is right, and only then is it allowed near production. The coding tool is also kept on a tight leash — when I ask it for information, I ask it to look and report only, changing nothing. The refresher I used to write this very article was exactly that: read-only, no edits, no risk.

This is the thing I have come to believe most strongly. AI did not remove the judgment from this work. It relocated it. I no longer spend my hours pushing pixels or typing code. I spend them deciding what good looks like and whether what came back meets it. The bottleneck moved from can I build this to do I know what I want — and the second is a far better place for a founder to be spending their time.

One honest caveat

A piece like this should admit what is not finished.

My platform currently has two visual generations living side by side: the new lime-and-charcoal system on everything built since mid-May, and an older grey-and-green look still carried by the legacy app and parts of the public site. Consolidating the old onto the new is real work that is not done yet. The new tools made it dramatically cheaper to build the future; they did not magically rewrite the past overnight. Anyone selling you a one-click total transformation is overselling. What you get is a step-change in pace, applied screen by screen, with your judgment in the loop the whole way.

Why this matters if you run a small business

Strip away the specifics and here is the general truth. Two of the most expensive, hardest-to-hire skills in software — visual design and front-end engineering — just became things a non-specialist founder can direct with a clear head and a couple of AI tools. The cost of looking professional, of shipping a polished interface, has collapsed. That collapse is not coming; for the work I have described, it has already happened.

For a small company, the implication is liberating and a little uncomfortable at once. The excuse “we can’t afford to make it look good” is gone. Which means the standard your customers quietly hold you to — does this look like a real product — is now one you can actually meet, solo.

What to do about it

If you run a product, a website, or any software your customers touch, there is one concrete thing worth doing this week. Pick a single screen that embarrasses you a little — the one you would not want a prospect to see first. Open a design tool, describe the better version in plain words, and refine it until it looks right. Then hand it to a coding tool to build, on a copy, where nothing can break. Ship that one screen. Measure whether it changes how the page performs or how people respond.

If it works, do the next screen. If it does not, you have lost an afternoon, not a quarter’s salary. That is the whole shift in a sentence: the cost of trying just fell far enough that trying is the obvious move.

The wall I lived behind for years is gone. The only question left is whether you walk through the gap.

A note on this article: it was written from a factual audit of my own platform’s code, and it is paired with a video produced inside MyAvatar’s own Movie Studio from this very text — built, fittingly, on the interface this article is about.

About the author

Mogens Rye is the founder of MyAvatar, an AI-native video production platform based in Copenhagen. He is a former stockbroker turned entrepreneur, and has built MyAvatar’s multi-engine platform as a partnership between a single human founder and a team of specialised AI agents. MyAvatar’s platform is available at myavatar.dk.