elizatrinity.tech launch series · Part 1

I Had 45 Pages of Sermon Notes in OneNote. Here's What I Did About It.

I am not a developer.

For most of my adult life, I worked in tech operations and account management. I knew how to use tools. I knew how to manage processes. But writing code from scratch? Not me.

So when I tell you I built an automated AI pipeline that transcribes sermons, generates structured study notes, and publishes them to a website — without touching the code manually — I want you to understand where that started.

It started with a preacher I couldn't keep up with.


The Problem

I've been listening to a teacher I follow for years. His teaching is different from most preachers I've encountered. He doesn't just motivate you. He takes you deeper into Scripture — verse by verse, layer by layer — and the revelation he carries comes fast.

I'd be listening and scrambling to write at the same time. Pause. Rewind. Write. Miss the next point. Pause again.

Over time I had built up 45 pages of notes in OneNote. Highlighted Bible verses in different colours. Timestamps. Partial sentences. It was a mess that only I could decode.

OneNote sermon notes page, showing handwritten-style highlighting, timestamps, and partial sentences from a sermon

I knew the content was valuable. I just couldn't access it in any organised way.


Initial Attempt: YouTube Transcript + ChatGPT

I found a browser extension that could pull the auto-generated YouTube transcript for any video. I'd copy the full transcript and paste it into ChatGPT.

That worked — until it didn't.

Free ChatGPT had a character limit. Long sermons got cut off. I was missing the second half of teachings. And every time I opened a new chat window, I had to re-explain the context from scratch. What format I wanted. What the site was for. What kind of reader I was writing for. Over and over again.

I built a master prompt and saved it so I could paste it at the start of every new conversation.

The ChatGPT conversation where the master study-note prompt was first created, so every future sermon could be generated in the same format

Even with a master prompt, the output was inconsistent. Sometimes bullet points. Sometimes numbered lists. Sometimes the tone shifted completely. I'd ask for a redo. Get another version. Still not right.


The Manual Publishing Workflow

When I did get output I was happy with, here's what the publishing process looked like:

  1. Copy ChatGPT output
  2. Open Wix editor
  3. Paste and manually reformat
  4. Go back to ChatGPT and ask for a header image idea
  5. Generate image separately
  6. Ask for alt text
  7. Ask for SEO keywords
  8. Back to Wix
  9. Add everything manually
  10. Publish

One article. Hours.

The old elizatrinity.com blog homepage, showing inconsistent formatting from the manual publishing era

I also built a master Google Sheets tracker to try to keep everything organised — tabs for skillset tracking, blog setup, pre-publishing checklists, monetisation tracking. It grew to eight tabs.

The master Google Sheets tracker used to manage the manual Wix blog workflow — skillset, blog setup, publishing checklist, and monetisation tabs

That spreadsheet tells you everything about how unsustainable the workflow was. When you need a spreadsheet that big just to manage one blog, something is wrong.

Here's what the actual copy-paste request looked like, mid-conversation:

A ChatGPT conversation showing the manual copy-paste workflow — asking for output formatted so it could be pasted straight into the blog


The Moment Everything Changed

It was actually a school assignment that cracked this open.

I was doing a course at Lithan — a tech upskilling programme here in Singapore — and one of the projects was an e-commerce AI application. We had to implement features like natural language search, personalised recommendations, and a chatbot.

I remember asking ChatGPT: "How could I apply these same AI features to my own faith content project?"

That's when it explained RAG — Retrieval-Augmented Generation. A system where AI doesn't just know general information, but can search and reason over your specific content. Your own knowledge base.

Something clicked.

What if the teachings became a searchable knowledge base? What if every sermon, every golden nugget, every scripture reference was indexed and retrievable? Not paraphrased — preserved.

That became the vision behind what is now Revealed Insight.

But between the vision and the reality was a wall.


The n8n Dead End

I tried to automate the workflow using n8n — a no-code automation tool. Spent three days troubleshooting with ChatGPT, trying to connect it to Wix and get an automation working end to end. Self-hosted n8n through Docker. Enabled Wix's Velo dev mode just to get API write access. Tested webhooks. Still couldn't get a single item to write reliably into the Wix CMS.

Enabling Velo Dev Mode on the Wix site editor, part of the failed attempt to connect n8n to Wix's CMS

Three days in, I gave up. I was stuck. I could see what I wanted to build. I just had no idea how to get there.


What Happened Next

That's where this story picks up in Part 2.

Within a few weeks, everything I had been trying to do manually for months was running automatically.

Revealed Insight went from zero to 30+ articles live in the time it used to take me to publish five.

(Article 2 — Building the Pipeline — linked here once it goes live.)


If this resonated — especially if you've ever sat with a notebook, pausing and rewinding a sermon trying to capture something too deep to catch in real time — I'd love to hear from you.