A post has been going around the instructional design corner of LinkedIn this week. Kyal Stahl described using Claude Cowork to build an entire course in Articulate Storyline in minutes — logical structure, technically accurate quiz questions, content aligned to objectives. He was blown away. Then he actually ran through the finished module and felt something was missing: no sense of the daily frustration of the job, no feel for team dynamics or the unwritten rules that make work hard. His conclusion: “AI can construct the skeleton. Only human instructional designers can give it a heart.”
I don’t disagree. I’ve made close to the same argument myself about what gets lost when you hand creative judgment fully to AI. Five hundred-plus reactions and dozens of comments on that post say the “AI courses feel cold” take landed, and it should — it’s true.
But it’s not the risk I’d lead with.
What actually happened when I tried it
I tried something close to what that post described: letting an AI agent drive Articulate Storyline directly, building slides and structure inside the actual authoring tool instead of just generating a script or outline for me to assemble by hand.
It didn’t just come out soulless. It messed up my Storyline project. I spent longer untangling what the agent had done to the file than I would have spent building the module myself.
I left a comment on that post at the time: this is a little dangerous, because tools like Claude Cowork have the kind of file-system access that can delete or overwrite things on your computer, not just misjudge a quiz question. I wasn’t being dramatic. I’d just watched an AI agent take real liberties with a live project file, on my own machine.
The real trap for 2026 isn’t cold content. It’s file-system access.
Every take on this story so far, including the original post, frames the risk in pedagogical terms — shallow content, missing nuance, an information dump wrapped in slick formatting. That’s real, and worth taking seriously. But it’s the visible risk. The quieter one is that the newer “agentic” AI tools aren’t just writing to a script or a document anymore — they can open, edit, and manipulate files and applications directly on your machine, with roughly the same permissions you have.
A course that feels emotionally flat is a rewrite. A corrupted Storyline project, an overwritten export, or a deleted asset folder is a much worse afternoon. That’s not a hypothetical for me — it’s what happened.
How to actually use this workflow, if you’re going to
I’m not telling anyone to avoid AI-assisted course building. I use AI in production every week, and I’ve written before about how well it fits into a Storyline pipeline when it’s used the right way. But if you’re going to let an agent operate your authoring tool directly instead of just generating content for you to review and assemble:
- Don’t run it against your primary machine. Use an isolated environment — a cloud desktop instance, a spare Mac mini, anything that isn’t the computer holding your client files and everything else you can’t afford to lose.
- Back up the project file first, every time — not just the first time you try it.
- Treat the output as a rough draft to inspect, not a finished module to trust. The same review discipline that catches “flat” content also catches broken timelines, misplaced triggers, and reordered layers.
- Read what permissions you’re actually granting. “It can build a course in minutes” and “it can read and write files on your computer” are the same sentence — only one of them makes it into the excited LinkedIn post.
Where I actually agree with the original post
The soul problem is real, and I’d rather see it solved than waved away. But the file-system problem is the one nobody’s screenshotting for engagement, and it’s the one that can cost you a project, not just a rewrite. If you’re experimenting with agentic AI tools driving your actual authoring software, do it somewhere expendable. DYOR, and then some.
If you want a second set of eyes on where AI genuinely helps versus where it’s not worth the risk in your production pipeline, contact us.





