After six years we have a new website. We built it with AI, and we want to tell you how

The site you knew ran on Gatsby for six years. A reasonable choice back then. But in six years almost everything changed: what we do, how we work, and what we even expect a website to be. We help companies put AI into real production, and yet our own site said nothing about that and got harder to maintain every year.

The design didn't start in Figma

This time we did not start with a designer and a blank canvas. I sat down with Claude Design and described what the site had to do: who we are, what we sell, how it should feel. Instead of a static mockup we got a clickable version of every page right away. Home, services, work, careers, blog. When something felt off, I said so in plain words and saw a new version a minute later.

It was not magic on the first try. We threw away the first few directions. But an iteration that used to take a day and a meeting now took minutes. That is the difference that is hard to explain until you have felt it once.

From design to code in days

We handed the chosen direction to Claude Code, which did most of the implementation. The site runs on Next.js, it is bilingual (Czech and English), and we manage content in our own lightweight CMS rather than a heavy system we would have to feed. Deployment is automatic: every merge builds itself and ships to the server.

What AI took off our hands, the work I used to click through for days:

• Project scaffolding, routing, and the bilingual content plumbing
• Building the individual pages and components from the design
• A custom CMS for the blog and careers on top of a database
• The whole deployment pipeline and server setup

Where it got hard (and why you still need an engineer)

Let's be honest here, because this part usually gets left out of the marketing. The last twenty percent is still real work. The animated industry strip on the homepage looped wrong and left an empty gap after the first pass. On the first CMS deploy the app came up but the database stayed empty, because the migrations never made it into the production image. Neither of those was a "build me a website" problem. Both were "do you understand what is happening under the hood" problems.

AI did not get rid of our engineers. It changed what they spend time on. Instead of writing boilerplate, we spend it on the decisions that matter.

What this means for you

This is exactly how we build client projects too. Not by having AI spit out a site and hoping. AI takes the boring, repetitive part; we own the architecture, the quality, and whether the thing actually holds up in production. That is how we ship working software in a fraction of the time it would have cost a few years ago.

This site is our own proof. We built it fast, in the open, and it runs the very blog post you are reading. If you are weighing something similar, get in touch. We are happy to show what it would look like for you.