Key Takeaways
- One prompt reformatted 48 lessons across a GoHighLevel membership course. Titles, headings, structure, the whole thing.
- The browser route was slow and fragile. Claude switched to the GHL API on its own after reading the docs, and that is what made it work.
- The first pass took around 30 minutes of watching. Round two ran in roughly 90 seconds, because the system already knew the job.
- This is a real GHL use case, not a demo: course build-out is one of the most tedious jobs in any membership site.
- Bay Area operators sitting on unfinished courses are usually blocked by formatting, not by content.
If you have ever built a course inside GoHighLevel Memberships, you know the actual work is not the teaching. It is sitting in the lesson editor for hours, pasting text, fixing headings, naming things consistently, and doing it 48 more times.
Michael Le at On Point Tech Solutions in Los Gatos handed that job to Claude and recorded the whole thing. Not a highlight reel. The full session, including the part where it did not work.
The First Attempt Went Through the Browser
The obvious approach is to let the AI drive the browser: open the lesson, click into the editor, type, save, move to the next one. That is what it started doing, and you can watch it work through the GoHighLevel interface in real time.
It works. It is also slow, and every click is a chance to break. Thirteen minutes in, Michael is still watching it figure things out.
Then It Read the API Docs
This is the turn in the video, and it is the part worth stealing. Rather than keep clicking, Claude went and read GoHighLevel's API documentation, worked out that the lessons could be written directly, and switched approach.
From there it validated all 48 lessons, refreshed its token, and pushed them. The result is a course where every lesson carries a real title, a clean H1, and a consistent structure: what the lesson covers, the story, the fix, what to keep.
The lesson for anyone automating GoHighLevel work is simple. When a tool feels slow through the interface, the question is whether the interface is the only door. On GHL it usually is not.
Round Two Took 90 Seconds
Michael ran it again on a second course. The first pass had taken about 30 minutes of sitting and watching. The second one was moving in roughly a minute and a half.
Nothing about the model changed. What changed is that the approach was already known, so none of the discovery had to happen twice. That is the actual compounding effect of building systems instead of running one-off prompts, and it is the same reason the second client build in any agency is faster than the first.
Why This Matters for Bay Area and San Jose Businesses
A lot of local businesses in San Jose and around the Bay Area are sitting on a course, an onboarding sequence or a client training library that is 80 percent written and 0 percent published. The blocker is almost never the expertise. It is that turning a folder of notes into 48 properly formatted lessons is a week of evenings nobody wants to spend.
That is exactly the shape of work that AI handles well: high volume, highly repetitive, with a clear right answer for what good looks like. It is also the kind of job that quietly never gets done otherwise, which means the cost is not an hourly rate. The cost is the product not existing.
For a service business, a finished course is a real asset. It is the thing you sell at the low end, the thing you hand a new client during onboarding, and the thing that lets you stop repeating the same explanation on every call.
Practical Steps
- Write your format once. Decide what a good lesson looks like, in detail, before you automate anything. The AI applies a standard, it does not invent one.
- Get your content into one place first. Automation cannot format what it cannot read.
- Let it try the API. If a tool has documented endpoints, that path is faster and far less fragile than clicking through the interface.
- Watch the first run all the way through. That is where you learn what your prompt actually left out.
- Expect the second run to be dramatically faster, and plan the rest of your build around that.
Watch the Full Session
Final Thoughts
Forty eight lessons is not the headline. The headline is that the job went from a week of evenings to one prompt and a coffee, and that the second time it ran it took 90 seconds because the system had already learned the route.
On Point Tech Solutions is a Go High Level consultant in Los Gatos serving San Jose and the wider Bay Area. We build the memberships, funnels and automations, and our clients own what we build. Done for you, or taught with an SOP so your team can run it themselves. Start at optechsol.llc.