How I Used GPT, Claude + Lovable to Build Halden

How I Used GPT, Claude + Lovable to Build Halden

This is the process behind my Lovable Built It for Small Business Challenge entry.

I didn’t start with:

“Build me a detailing website.”

I started with the actual challenge brief, the business problem, and the experience I wanted to create.

I pasted the requirements into GPT, developed the concept, refined the copy and customer flow, and then moved the build into Lovable.

From there, the process became a constant loop.

Build → inspect → copy the site back into GPT or Claude → critique it → improve the prompt → rebuild.

I would often copy nearly every word from the site back into the models and ask what was missing, confusing, inconsistent, or not aligned with the original vision.

I also researched what other service businesses and software companies were doing — current customer experiences, automation, checkout patterns, and where agentic commerce appears to be heading — and brought those ideas back into the project.

That research helped push Halden beyond a normal appointment website into something designed for both people and AI assistants.

The biggest lesson from the process:

Don’t stop when the build works. Analyze it again.

Even after the product looks finished, run it back through the models, test the assumptions, refine the experience, and keep iterating.

That back-and-forth between human direction, AI analysis, research, and Lovable implementation is how I built Halden.

Halden Detail Co. was built in Lovable for the Built It for Small Business Challenge.

The challenge is centered on reducing the friction between an inbound customer inquiry and a confirmed appointment.

Halden handles that booking process while also exploring what happens when an AI assistant becomes another interface into the same business.

@Lovable · #LovableChallenge

Jason T Wade is the founder of BackTier and works on AI Visibility, Generative Engine Optimization, Answer Engine Optimization, entity architecture, and emerging agentic experiences.

His work focuses on how businesses can become easier for AI systems to discover, understand, recommend, and interact with.

Halden / Lovable Build
https://winnerjasonwade.lovable.app

Lovable
https://lovable.dev

Jason T Wade
https://jasonwade.com

BackTier
https://backtier.com

Email
email@jasonwade.com