My role
I designed and built these systems to help the team create content at scale, with human approval at every stage.
The problem
A small team supported courses, in-app guides, and videos for a growing product. Courses sometimes took six weeks or more from intake to publish, with a lot of hand-holding to build each resource’s HTML. Reviews kept catching the same mechanical errors instead of judging substance.
What I did
- Found the pattern first. I analyzed revision reports and found the same handful of issues on almost every course review. Style rules and HTML guidelines weren’t being followed consistently while drafting.
- Turned rules into checks. Style and format rules became deterministic lint checks that run before any human review.
- Built pipelines with approval gates:
- A course builder covering intake, research, outline with a rigor check, drafting, shared review, automated QA, LMS formatting, and lessons learned.
- An enablement version for internal training and sales enablement.
- An in-app pipeline that turns support signals into ranked opportunities, drafted guides, and Figma frames.
- Kept people in charge. None of the pipelines publishes or sends anything without clear approval.
Results
- Course turnaround dropped from six or more weeks to three weeks or less — about twice as fast.
- At peak, four contributors were producing courses through the pipeline.
- 25 self-paced courses, 15 how-to videos, and 120+ in-app guides since April 2026.
- Reviews shifted from formatting to accuracy, voice, and instructional quality.
What I learned
AI drafting is only as reliable as the standards it’s held to. Making those standards explicit and testable did more for quality than better prompts.