All articles
Company News & Market Insights

AI Did 80% of Our Website in Days. The Other 20% Took 5 Weeks

Stop guessing which AI tools actually work. We rebuilt iGsuite - The Ultimate Affiliate Software from scratch using Claude, Lovable, and Next.js - here's the honest breakdown of what delivered and what cost us five weeks.

June 27, 2026 4 min read
AI Did 80% of Our Website in Days. The Other 20% Took 5 Weeks

The brief was simple: migrate off WordPress, go live on something modern. Done.

It was not done.

Our co-founder and senior software engineer Valentina led the iGsuite’s website rebuild from architecture to DNS cutover - and she used AI at every stage. Claude Code. Lovable. ChatGPT. Midjourney. Not as a gimmick, but as actual build infrastructure. The result is the site you see at igsuite.com today - dark, cinematic, glassmorphism, premium. The kind of landing page where you feel in the first three seconds that the people behind it know iGaming.

We're sharing this because building a credible web presence is one of the first real challenges any new iGaming (or any, really) business faces - and the landscape for how you do that has fundamentally changed. If you're a startup operator, a new platform, or a team about to migrate off a legacy CMS, this is the honest account we wish we'd had before we started: what the AI tools actually did, where they fell short, and what it takes to go from vision to live in 2026.

Here is an honest account of how we got there, what the AI tools actually did, and what the process taught us about building with AI in 2026.

The Vision: Premium B2B, Not Generic SaaS

Before a single line of code, the brief was specific. Dark, cinematic. Glassmorphism - frosted glass navigation, cards lit from inside. Electric blue as the primary accent, warm gold #e8a33d for stars and highlights, brand red used almost never. IBM Plex Sans for headlines, IBM Plex Mono for the eyebrow labels. Restrained - no disco colors, no neon overload. The kind of site that communicates immediately: this was built by operators for operators.

That vision took five weeks to execute, not five days. And the gap between the two is the most instructive part of the story.

How the AI Stack Was Actually Used

  • Claude Code handled the architecture. Next.js + Sanity CMS. Blog schema, SEO scaffolding, deploy pipeline - all of it built in the first few days. The bones were solid fast. Claude Code is genuinely strong when the brief is precise: ‘build a Sanity webhook that triggers a redeploy on publish’ produces clean results. ‘Make it feel more premium’ does not.
  • Lovable handled design prototyping. The About page - flip cards, ambient gradients, the full dark cinematic treatment - was built properly in Lovable first, then exported and ported into Next.js. The tool gets you 80% of a polished visual in minutes. Matching the remaining 20% pixel-for-pixel in production took the rest of the project.
  • ChatGPT and Midjourney were tried and dropped. Both were tested for ambient background generation. Both kept producing full hero compositions with text baked in. Neither matched the brand kit cleanly. Sometimes the right answer is to recognize when a tool isn't the right fit for a specific task and move on.

What No AI Tool Could Do: Learn Your Brand

This is the lesson at the center of everything. No AI tool — not one — understood the iGsuite brand on the first prompt, or the tenth. The navigation went back and forth ten times before settling. The glow on cards had to be specified precisely: top edge, not corner. The rule that only one keyword per heading gets the blue accent - that came from iteration, not instruction. The exact blur radius that reads as 'frosted glass' rather than 'blurry rectangle' - same.

Every font, every color decision, every nuance was manually worked through and then had to be repeated until it landed. AI tools compress timelines. They do not replace taste, and they do not absorb your brand identity from a single brief.

The formula, honestly: AI gives you 80% in 5% of the time. The remaining 20% takes the other 95%.

Key Takeaways

For operators and teams considering an AI-assisted build:

  • Precision beats volume in AI prompts. Vague briefs produce generic output. The more specific the instruction — component names, exact colors, behavior on hover — the more useful the result. Before each task, map what you want to the actual file or component it lives in.
  • Prototype visually before you build. Use Lovable or a similar tool to reach 80% of the visual target before committing to production code. Porting a finished design is faster than designing directly in code.
  • "Connected" is not the same as "controls." Our Sanity CMS was wired before the site was hardcoded. The integration existed; it just didn't do anything. Verify that your CMS actually drives the live content before you consider it done.
  • Mobile is its own project. A site that reads as cinematic on a monitor can be a lagging, white-flashing mess on a phone. Budget dedicated time for mobile performance — it will take longer than everything else combined.
  • Run an AI search audit of your brand. Ask the major AI tools what they recommend in your category. If you don't appear in that answer, you have a plan to build.

The site is live. The glass nav scrolls smooth. The forms work. The blog is running real content. And the next project - making sure iGsuite shows up when it matters - has already started.

And if you're at this stage yourself - congratulations, the hard part is done. The next chapter is visibility: SEO to build organic authority, and GEO (Generative Engine Optimization) to make sure your brand shows up when AI tools answer the questions your future customers are already asking.

Go see it at igsuite.com.

🚀 Let's Scale Together: We're building the future of the iGaming ecosystem, and we want you in the loop. Follow us on LinkedIn for a behind-the-scenes look at our newest features.

Book a demo