A few weeks ago, I fed Claude a PDF of an existing online course and asked it to make the content more engaging. Simple request. I expected a simple win.
The first output looked great on the surface with livelier language, more interactivity, better flow. But the more I dug into it, the more the cracks showed. It had traded structure for polish. The instructional design wasn’t there. Not because the model couldn’t do it, but because I’d never asked for it.
What I actually asked for the first time
“Make this more engaging” is a vibe, not a brief. It’s the kind of instruction that sounds specific but leaves almost everything important undefined:
- Engaging for whom?
- Engaging within what constraints – what can the platform actually support?
- Engaging in what tone – corporate, casual, technical, playful?
- Engaging without losing what?
I hadn’t answered any of that. So the model filled in the gaps with its best guess, and its best guess was generic.
What changed the second time
I rebuilt the prompt the way I’d build a design brief for a real project:
- Audience – Who’s taking this course, what is their role, their prior knowledge, the context they’re learning in?
- Platform constraints – Which LMS or authoring tool this would live in, so “engaging” meant interactions that could actually be built, not ideas that would die in production.
- Tone – Matched to the audience, not a generic idea of “engaging” language.
- Content boundaries – Keep the original substance intact. Improve the design. Don’t rewrite the meaning.
Same source PDF and a prompt that used instructional design principles resulted in a completely different output. Why? Because this time the prompt did the job of an actual design document, not just a request.
The lesson
The model didn’t get smarter between the first prompt and the second. My prompt got more like a design brief.
That’s the part I think gets missed in a lot of AI conversations right now. People treat prompting like a search bar. They type a request, get an output, move on. But the prompts that actually produce good work aren’t shortcuts. They’re doing the same job instructional designers have always done:
- Define the audience before you define the solution
- Understand the constraints of where the work will actually live
- Set a clear tone and success criteria up front
- Protect the substance while you improve the design
- When the output misses, diagnose why. Is it because vague instruction was used, missing context, wrong assumption?
None of this is a new skill. It’s the same instinct instructional designers already bring to every project but just aimed at a different kind of collaborator.
Prompting isn’t a shortcut around design thinking. It’s design thinking, with a faster feedback loop.
The people who get the most out of AI in L&D aren’t choosing between good templates and good thinking but they are using both. A solid prompt template saves time and keeps you consistent, the same way a course template does.
But a template is a starting structure, not a substitute for the thinking behind it. It works best in the hands of someone who already knows what to adjust and why — who’s the audience, what are the constraints, what does success look like.
Prompting isn’t a shortcut around design thinking. It’s design thinking, with a faster feedback loop. And a good template is just that thinking, packaged so you don’t start from zero.
Want to learn more on how to prompt for L&D design? Join our course, The L&D Professional’s Guide to AI Fluency, and learn how to prompt using a framework and become more AI fluent in your role.
