Landing Page Generator
QR codes that don't break, landing pages that write themselves. Built it for a client, now it's becoming its own thing.
The Challenge
The Problem
David DeDiego's building a community-focused passion project and needs landing pages for different audiences. Lots of them.
Why I Took This On
David's doing something real for his community, not chasing a quick buck. I wanted to help with that. Plus, the challenge was fascinating.
Constraints
Had to be dead simple—if they could use Canva, they could use this. Also needed to generate clean, fast-loading pages without any bloat.
The Process
Initial Approach
Started with the end user in mind: someone who's good at their trade but doesn't want to think about code. Built a component library of pre-styled blocks they could drag into place, then wired up one-click publishing to a custom hosting setup.
What Went Wrong
There were several things that blew up the simple concept. Initially, there was only one kind of landing page available - a three-point layout that was supposed to transition into a scored quiz. However, the quiz wasn't always wanted or needed. And in fact, there needed to be several archetypes. Back to the drawing board and have enough user input so that the AI could determine the best type of landing page as well as the content for it. Time for new fields and new questions.
Breakthroughs
n8n happened. When I realized that I could "un blackbox" the AI process and break things into discrete steps, everything changed. Instead of an impenetrable process happening that was diffrerent every time, I was able to break the entire process into discrete steps. Decide where and when to leverage AI in the process.
The Solution
n8n became the place for my wireframed idea to come to life. Instead of handing off my input to an AI with some instructions and hoping for the best, I could mix and match.
Create a form -> Fill out the form -> Take those values and feed it to an AI copywriter -> AI proofreader reads, takes notes, and scores the output. If it's a passing score, then off to a human approval queue it goes! Not meeting the standard? Notes, score and input go back to the copywriter node for another go-round.
At the end, the approved output goes into a queue for manager approval and on they go.
Two wins: one, QR codes never go bad. Instead of pointing to "Christmas 2022" or some other expired promotion, any QR codes out there on old flyers, posters, or forgotten web pages will still point at the default welcome page for any given category. And two - and arguably more important - the people who work on promotions aren't required to be website experts in order to do their jobs!
What I Learned
Skills Gained
Unexpected Discoveries
In seeing the n8n layout spelled out so clearly, I understood how I'd be applying this to so many areas in my computer life. Automate easy stuff? Got it! Get really complex? Oh yeah! Break things up into sub functions? Totally makes sense!
What I'd Do Differently
Not too much. It was a feel my way through the dark until I find the light switch situation. No way to adequately prep for that ahead of time.
Where It Stands Now
Current State
It's evolving. It's in use for making basic pages. Soon it will be a fully fledged tool on its own! Likely it will be an external call after awhile.
What's Next
This is going to be turned into an external tool that is self-contained. It's too valuable to not package for other people to use.
The Bigger Picture
This will help with everyone that has a business and needs to communicate with potential and existing customers in a consistent professional way. I can't wait to build the robustness and flexibility into it that this idea deserves.
The Impact
Client Results
ChurchScroll.com
Personal Impact
I'm so excited, not only by the breakthrough and figuring this out, but also the insight around how this was going to apply to everything else in my head that was going to need to be automated in the future! Programmatic where it needs to be the same every time and generative in the spots where it makes sense!