Build a course with AI only where it helps: human outcome and outline, AI for examples, analogies, copy and voice
A course creator with over twenty years of experience describes keeping course design human and using AI for examples, analogies, marketing copy and voice matching.
Evidence: The author reports this. We have not checked it beyond reading the source.
The business problem
An expert building a course wants faster examples and marketing material without the generic, invented content AI tends to produce.
What was tried
Define a measurable learner outcome, the actions needed to reach it and the obstacles to each action, without AI. This becomes the outline. Give AI existing examples and ask for broader ones that people in other industries would understand, usually about ten options, and pick the best. Ask AI to diagnose why an analogy does not fit and refine it. For marketing copy, have AI interview you first, supply copywriting frameworks, forbid invented details, and run a custom prompt that strips common AI patterns. Use a voice-analysis method so output drifts toward your own voice. For a course made for an employer, he outlined, dictated each explanation, and had AI transcribe and polish it. Test with real learners.
What was reported (not reported)
He reports faster idea testing, faster first drafts of marketing copy and useful feedback on analogies that would take days to reflect on. No sales, completion or time figures are given.
Limitations
An interview in a curated series with no metrics, and the tools are mostly unnamed. He says AI swings between useful insight and fabricated content, produces generic copy without guidance, and that the cleanup prompt never catches everything, so manual editing remains. He suggests roughly 90 percent of content stays human-written. Simulating a student with AI is something he has not tried.
What you need
Real expertise in the subject, an AI assistant, and a way to test the course with actual learners.
Sources
- AI Workflow Stories on Substack (interview with Rodney Daut) ↗ Firsthand write-up, published August 21, 2026
Source published: August 21, 2026. Last reviewed here: October 13, 2026. Spot a mistake? Tell us.
Related workflows
Build a small SaaS with Lovable, Supabase and Stripe using a living spec file and a habit of checking the database yourself
A founder builds with Lovable, keeps one living spec file the model reads before every feature, ships one feature at a time, and audits the database, logs and billing after each one because the tool reports success without checking.
Rebuild a company website in about three weeks with Claude for planning and design and Claude Code for the build
A marketer used Claude to plan and design each page and Claude Code to build and update a hard-coded company site himself, alongside his regular job.
Build a privacy-first Claude usage tracker as a Chrome and VS Code extension with Claude Code and publish it to three marketplaces
A small agency describes building an extension that shows Claude.ai usage limits, using Claude Code for the architecture, code, build setup, icons and store listing text, then publishing it to the Chrome Web Store, VS Code Marketplace and Open VSX.
Ship a Chrome extension update in under three hours and about eight dollars of tokens with Claude Code
A solo developer used Claude Code to release two versions of a small Chrome extension in two days, including debugging, refactoring, release notes and the store listing text.