I didn't make eight animated lessons in one night by asking AI to "make eight animated lessons."
I built a process that moved each chapter through production and review. This was a project with an online course platform. Eight chapters needed a full set of deliverables, and the complete version ran overnight.
I'm Young. I've worked across healthcare, education, and long-term care for 20 years. Over the past two years, I've used AI while running more than 15 projects at once. I've written about how I manage those projects and how I divide work among AI agents.
What eight chapters meant
Each chapter needed a finished video without subtitles, an internal review version with subtitles, a complete audio track, a transcript, a quality check record, and a review from the point of view of someone watching it for the first time.
Making each chapter by hand would have taken weeks. My job was to turn the repeated work into a process that could finish eight chapters in one night while still finding errors.
The gates inside the production line
The work moved through distinct stages:
- A storyboard had to exist before voiceover. The process blocked the next stage if it was missing.
- Each finished chapter went to a blind first-viewer review. That reviewer looked for covered text, false endings, black frames, missing frames at joins, and arrows that pointed the wrong way.
- A chapter with a problem went back for rework. The rest of the batch didn't get a free pass because one chapter looked good.
The first blind review found 16 problems that night. None was caught by the automated tests. The viewer check found them all.
That is why I designed the review into the process. Speed alone would have produced eight chapters with defects I hadn't seen.
The mistake I made with parallel agents
At one point I tried sending five AI agents to work in parallel. Nobody coordinated them. I had no condition for bringing their work back together. They spent time running and delivered nothing.
The number of agents was not the measure of capacity. Coordination and finishing were.
So for the animation work, I chose to complete and check each chapter in sequence. The same one-person-plus-AI structure appears in my medical security delivery and AI evaluation platform project.
What a client gets from this approach
| A common way to use AI | How I built this process |
|---|---|
| Ask AI to do one task | Set up a repeatable sequence of tasks |
| Move fast and hope the output is good | Put a check after each stage |
| Start many agents at once | Finish and verify each piece |
| Discover problems after delivery | Let the process flag work for rework |
This is how I handle large amounts of repeated work across more than 15 projects. AI does the production. I design the stages and review the results.
Questions I get
Does this only work for animated courses?
No. Stages, gates, blind review, and rework can also help with reports, batch content, and data processing.
Can one night of production still be checked?
That night, the blind review found 16 problems. Those chapters went back for rework before release.
Do I need deep technical skills to use this idea?
You need to understand the work well enough to define its stages and decide what a failed check should stop. The tools will change.
I'm Young. I've spent 20 years working across healthcare, education, and long-term care. The course platform in this story remains anonymous. If you have a large, repetitive job where quality matters, book a free conversation with me.