I think instructional designers should be designing again.

I built my first web course in 1999, by hand, because that’s how you did it. When SCORM came along, I believed in what it was trying to do: separate the course from the system that delivered it.

Build something once. Put it anywhere. Keep it. Reuse it.

That was a very good idea. I spent a good part of the next 20 years helping make it work.

I founded JCA Solutions, where we built SCORM engines, packagers, adapters, and a lot of the unglamorous plumbing underneath online learning. I became one of nine certified SCORM auditors. Over the years we built learning technology for organizations including the U.S. Army, U.S. Navy, Northrop Grumman, Stanford, LexisNexis, and FlightSafety.

So I’m not standing outside the authoring-tool industry throwing rocks at it. I helped build the world that made those tools necessary. And they were necessary.

Authoring tools solved a real problem

Building for the web by hand was hard. Most instructional designers weren’t developers, nor should they have needed to be. Authoring tools gave people a way to create online learning without worrying about JavaScript, browser compatibility, SCORM calls, manifests, APIs, and all the other technical work happening underneath the course.

That was progress. But there was a tradeoff.

The tools made courses easier to build by deciding, in advance, what you could build.

• Here are the layouts.
• Here are the interactions.
• Here are the transitions.
• Here are the things you can change.
• Here are the things you can’t.

For a long time that was a perfectly reasonable trade, because the alternative was learning to code.

I don’t think that’s the alternative anymore. AI changed it.

I saw the shift in my own work first

I spent 30 years writing code by hand. I was good at it, and for a long time that was simply how the work got done. Then AI came along, and I started using it to do more and more of the actual coding.

Not because I forgot how to code. And not because the experience I built over those 30 years suddenly stopped mattering. Quite the opposite.

I found myself spending less time writing every line myself and more time deciding what needed to be built, how it should work, and whether what AI produced was actually right.

And that got me thinking about instructional designers. They’ve spent years becoming experts not only in learning and design, but in the mechanics of the tools they use to create it. Building slides. Setting triggers. Managing timelines. Formatting screens. Previewing. Publishing. Going back and changing something. Publishing again.

That’s a lot of time spent operating software.

If AI can take on more of those mechanics, what happens to all that expertise? I think it gets redirected toward the part that mattered all along: designing the learning experience.

That realization became a big part of why we built coursetemplates.ai.

But there’s a problem with asking AI to build a course from scratch

I’ve tried it. AI will happily make you something that looks like a course. That doesn’t mean it’s a good course. And it definitely doesn’t mean everything underneath it works the way it should.

A course isn’t just what you see on the screen. There’s LMS communication. Completion. Scoring. Bookmarking. Responsive behavior. Accessibility. Packaging. All the things learners never see but that have to work.

We shouldn’t be asking AI to reinvent those things every time someone wants to build a course. So we built them already.

That’s really the idea behind coursetemplates.ai. We build the working foundation. You use AI to make it yours.

Start with something that works, then tell AI what you want

coursetemplates.ai is a library of working course templates, interactions, simulations, games and other learning experiences, designed from the beginning to be edited with AI.

You don’t start with a blank prompt and hope for the best. You start with something that already works. Then you tell AI what you want to do with it.

The template gives AI a solid foundation to work from. AI handles the implementation. You make the decisions. And that’s an important distinction.

I don’t think AI makes expertise less valuable

I think it makes expertise more valuable.

AI can produce a lot, very quickly. What it can’t reliably do is know whether what it produced is actually good.

• It doesn’t know your learners the way you do.
• It doesn’t know that an example is technically correct but completely wrong for the people taking the course.
• It doesn’t know when an interaction looks impressive but teaches absolutely nothing.
• It doesn’t know when the right answer is to cut half the course.

You do. That’s the expertise.

For years we’ve mixed that expertise together with proficiency in a particular piece of software. Knowing how to build a trigger became tangled up with knowing when you should use one. Knowing how to create an interaction became tangled up with knowing what learners actually need to practice.

Those aren’t the same skills. AI gives us a chance to separate them again.

The model we’ve ended up with is pretty simple

And it opens up a much bigger canvas

This is the part I’m most excited about.

An authoring tool can only give you features someone has already decided to put into the authoring tool. The web doesn’t have that limitation.

Today’s browser can handle 3D environments, animation, simulations, games, real-time audio, interactive storytelling, data visualization, complex interfaces, and things I couldn’t have imagined when I built that first web course in 1999.

For most instructional designers, much of that hasn’t been realistically accessible. You might have a great idea for a learning experience and immediately have to translate it into how could I make something like that in my authoring tool? Or worse, what’s the closest thing my authoring tool can do?

AI changes that.

• You don’t have to know how to program a 3D environment to have an idea for one.
• You don’t have to know JavaScript to decide that a learner should practice a conversation instead of answering another multiple-choice question.
• You don’t have to understand the code behind an interaction to know what you want that interaction to do.

That’s what interests me about AI in learning. Not that we can make the same courses faster. That we can make things we couldn’t realistically make before.

That’s why the templates matter

If AI can build almost anything, it can be tempting to think we don’t need templates anymore. I think the opposite is true. The more freedom AI gives us, the more useful a good starting point becomes.

A template shouldn’t tell you what your finished course has to look like. It should give you the parts that are already solved so you don’t have to solve them again. The underlying functionality works. The LMS communication works. The structure is there. The files are organized so AI can understand them.

Then you can change the part that actually should be different: the learning experience.

Not finished courses you have to squeeze your content into. Starting points you can take somewhere else.

I also don’t want to replace one kind of lock-in with another

This matters to me. I spent too much of my career working on interoperability to build another system that requires you to keep paying us just so something you created continues to work.

So the output is intentionally ordinary. HTML. CSS. JavaScript. Files. You download them. You have them. Your course doesn’t need to call coursetemplates.ai to ask permission to run.

If we disappeared tomorrow, what you built should still work. You should be able to put it in an LMS, host it somewhere else, hand it to a client, open the files yourself, or give them to another AI ten years from now.

Your course should belong to you. That was part of the promise behind SCORM in the first place. Twenty-five years later, I think we finally have the technology to take that idea much further.

Everything we build follows three principles

That’s also why we separate the engine from the content

There are parts of a course you want to change constantly: the words, the questions, the images, the colors, the scenarios, the examples. And there are parts you generally don’t: the underlying logic, tracking, progress, scoring — the machinery that makes everything work.

So we separate them.

That makes the course easier for people to understand, easier for AI to edit, and much harder to accidentally break. You shouldn’t need to understand LMS communication because you want to change a question. And AI shouldn’t need to rebuild the course engine because you asked it to make the buttons blue.

So that’s why we built coursetemplates.ai

Not because authoring tools were a mistake. They weren’t. They solved the problem we had. But I don’t think we have the same problem anymore.

For most of the history of eLearning there were basically two choices: use an authoring tool, or learn to code. AI introduced another one.

Start with something technically sound. Tell AI what you want to create. Use your expertise to make it good.

That’s what coursetemplates.ai is for. We’re not trying to tell instructional designers how to design. We’re trying to give them more room to do it.

Because after 25 years of building learning technology, I think the most interesting question we can ask right now isn’t “what will my authoring tool let me build?”

“What do I want to create?”

That’s a much better place to start.

— Brian M. Caudill