The core idea
Claude works best as a thinking accelerant, not a content vending machine. The faster you give it real constraints and push back on generic answers, the more useful it becomes. Everything in the guide is tested against real instructional design work — needs analyses, branching scenarios, evaluation plans, code-built learning tools.
What’s in the guide
- What is Claude? — the mental model, what it can and can’t do, how it differs from search
- Skills — slash commands and saved workflows, and how to build your own for L&D tasks
- MCP & Tools — how Model Context Protocol connects Claude to files, systems, and APIs
- Build with AI — a catalog of L&D artifacts Claude can produce, from storyboards to interactive HTML simulations
- Start Here — a five-day ramp plan for getting real value on actual work
If your team is starting from zero
Begin with the Start Here tab, not the capability catalog. The pattern that works: pick one live project, bring Claude a real artifact from it (an SME transcript, a draft storyboard, an evaluation plan), and ask for critique before generation. Teams that start by generating content get generic output and conclude AI is overhyped; teams that start with their own material learn where it’s genuinely strong in a week.