What if your next website didn't start with a design agency, a backlog of DevOps tasks, or weeks of back-and-forth with developers? What if you just… told an AI what you wanted, and it built it?

That's exactly what we did at Breathe IT. We used an AI assistant as our development partner to build the very website you're reading this on. Here's how it went - the good, the surprising, and the lessons learned.

Why we went this route

We're a Microsoft technology consultancy. We help businesses transform with Power Platform, Dynamics 365, AI, and Power BI. We live and breathe digital innovation - so when it came to our own website, we wanted to practice what we preach.

The traditional route - hire an agency, wait weeks for wireframes, go through revision cycles - felt slow and disconnected from how we actually work. We wanted something faster, more iterative, and more hands-on.

Can we build a professional website by collaborating with an AI assistant?

Spoiler: Yes. And it was way more effective than we expected.

The setup: Human + AI working together

The process wasn't "type a prompt and get a website." It was a genuine collaboration - more like pair programming with a very fast, very patient colleague.

1. We started with a conversation

No wireframes. No spec documents. We described what we wanted in plain language:

  • Who we are (Microsoft consultants based in Norway)
  • What we do (Power Platform, D365, AI, BI, Fabric)
  • Who we're talking to (IT leaders, decision-makers in Norwegian businesses)
  • The vibe we wanted (professional but human, innovative but grounded)

The AI assistant turned that into a site structure within minutes.

2. Iterative building in real-time

Instead of waiting days for a developer to implement changes, we worked in real-time:

  • "Move the services section above the about section"
  • "Make the hero text punchier"
  • "Add a section about our Microsoft partnership"
  • "The contact form needs a project description field"

Each change happened instantly. We could see the result, give feedback, and refine - all in the same conversation.

3. Content generation + refinement

The AI assistant drafted all the copy - page descriptions, service summaries, team bios, CTAs. We reviewed, adjusted the tone, added our specific expertise areas, and refined. What would normally take days of writing took hours.

4. Technical implementation

Under the hood, the AI handled a wide range of technical tasks:

  • Responsive design - fully adaptive across desktop, tablet, and mobile
  • SEO optimization - meta tags, Open Graph/Twitter Cards, semantic HTML, structured data (JSON-LD), sitemap.xml, canonical URLs, and hreflang tags for multi-language support
  • Performance - clean, minified code, lazy loading images, optimized asset delivery
  • Accessibility - proper heading hierarchy, alt texts, ARIA labels, keyboard navigation, color contrast compliance
  • Multi-language support - full Norwegian, English, and Swedish versions with proper locale handling
  • Cookie consent - GDPR-compliant cookie banner and consent management
  • Contact forms - secure form handling with spam protection
  • Analytics-ready - structured for easy integration with tracking and monitoring tools

We also established a full development pipeline using Azure DevOps - from task planning and version control to automated deployments. Every change goes through a structured process: task creation, development, review, and deployment. This means we can ship updates quickly while maintaining quality and traceability.

What surprised us

Speed

From first conversation to live website: days, not weeks. The traditional process would have taken 4–6 weeks minimum. We were iterating on a working site within the first session.

Quality of first drafts

The AI's initial output was already 70–80% there. Most of the work was refinement and adding our personality - not fixing fundamental issues.

It understood context

Once we explained our business, the AI assistant remembered and applied that context throughout. Service descriptions connected to our expertise. The tone stayed consistent. It felt cohesive without us micromanaging every detail.

We still needed human judgment

The AI couldn't tell us what to say - only how to say it. Strategic decisions about positioning, which services to highlight, what makes us different - that's still us. And it should be.

What we learned

AI is a multiplier, not a replacement

The AI assistant made us 10x faster, but it didn't replace our thinking. We still defined the strategy, made creative decisions, and ensured everything reflected who we actually are.

Plain language beats technical specs

We got better results describing what we wanted in everyday language than writing detailed technical requirements. "Make it feel more welcoming" worked better than "increase padding by 20px and change the font weight."

Iterate fast, ship faster

Because changes were instant, we could try things we'd never bother requesting from a developer. "What if we flip the layout?" - done in seconds. This freedom led to better outcomes.

Keep the human in the loop

Every piece of content got a human review. The AI is great at generating, but the brand voice, the specific expertise - that's ours to own.

Key takeaways

  • ✅ AI assistants can build production-quality websites - not just prototypes
  • ✅ The process is conversational, not technical - describe what you want, refine what you get
  • ✅ Speed is the biggest win - days instead of weeks, with more iteration cycles
  • ✅ Human judgment stays essential - strategy, brand voice, and quality control are still yours
  • ✅ We practice what we preach - at Breathe IT, we don't just recommend AI; we use it ourselves

Read also:

2026 Kick-off: New Office, New Colleagues, New Energy

We've opened the doors to our new space in Stavern, welcomed three new colleagues and an AI, and kicked off 2026 with workshops and customer sessions.

Want to build something with AI?

We help businesses harness Microsoft technologies and AI to work smarter, move faster, and deliver more.

Talk to us →