The Green Belt Project

Can AI build a 218-lesson bilingual Green Belt curriculum in just a few weeks?

I decided to find out. But two weeks in, the experiment hit a wall. When I used AI to consolidate the global standards (ASQ, IASSC, CSSC, LSSA), the output became too broad and lost its edge.

That’s when it stopped being a tech experiment and became a process experiment.

I realized that scaling with AI works the same way as scaling improvement on the shop floor: quality and progress drop when the steps are too big, and the oversight is too loose.

So I changed the approach:

• I designed the core structure myself instead of letting AI decide.
• I used VBA to break the work into small, iterative chunks for the LLM’s.
• I curated every AI-generated excerpt before generating complete lessons, keeping quality under control.
• I automated the repetitive work (formatting, HTML, CSV imports) so I could focus on curation and quality.

The result is The Green Belt Project — now live and fully open-access.

But the real outcome wasn’t the website. It was the confirmation that AI doesn’t replace domain expertise — it makes the expert’s ability to structure the work, set constraints, and guard quality even more critical.