What an Instructional Design Founder Actually Builds

A friend of mine, a very decent golfer, spent a fortune on a new driver. The marketing promised twenty more yards, a lower spin rate, the kind of technological advantage that would supposedly transform his game. For the first few rounds, he loved it. He did, occasionally, hit it twenty yards further. Usually deep into the trees to the right of the fairway.

His scores didn't change. If anything, they got a little worse. The new tool, powerful as it was, only amplified a flaw in his swing. What he needed wasn't a more forgiving clubhead; He needed better course management, a more reliable 5-iron, the patience to play for the centre of the green instead of hunting for the flag.

It strikes me as a funny parallel that the world of AI course creation is full of new, shiny drivers. Every other tool promises to build a course faster, to generate more content in less time, to give you those extra twenty yards. And much like my friend’s driver, they are powerful. You can, with a simple prompt, generate thousands of words of plausible-looking training content in seconds.

But they often help you hit the ball into the trees, faster.

More than a thousand courses later, the lesson is that the hard part of corporate L&D isn’t the initial creation. The blank page is a challenge, but it’s a finite one. The real, grinding, expensive problem is stewardship. It's what happens after the course is built. It's course management, not just the tee shot.

The seduction of the blank page

When you’re building a software product, the easiest thing to demo is the 'magic trick'. You type a sentence, click a button, and a whole course appears. It looks fantastic. It feels like progress. It’s what venture capitalists get excited about and it’s what makes for a great two-minute marketing video.

This is the equivalent of the launch monitor at the golf shop. Pure numbers, perfect conditions, no consequences. The blank page is seductive because it offers a clean, contained problem: turn this idea into a thing. Almost every AI tool in the learning space is optimised for this moment.

But it's a fiction. In the organisations I’ve worked with for two decades - the healthcare trusts, the financial services firms, the national charities - the starting point is almost never a genuinely blank page. It’s a messy folder of existing materials. It’s a forty-module library of compliance training that is slowly, expensively, drifting out of date. It’s the director of learning who has inherited a suite of courses and has no idea which ones are still fit for purpose.

Take the scenario of a designed piece of mandatory training for a legal practice. Let's say it cost them about £15,000 all-in. It was a great course, grounded in specific scenarios. Six months later, the regulator tweaked a single point of guidance. Suddenly, their expensive, well-designed course wasn't just out of date; it was actively giving bad advice. It had become a risk. The cost to unpick the relevant sections, update the content, re-shoot a video, and re-test the assessments was almost a third of the original budget. That is the reality. The problem isn’t the blank page. The problem is the half-life of knowledge.

Building for the second shot, not just the tee shot

In golf, they say you 'drive for show and putt for dough'.  Your tee shot gets the applause, but your score is made in the next hundred and fifty yards. It's made in the awkward lies, the recovery shots, the strategic lay-ups. The same is true of learning design.

Creating the first version of a course is the tee shot. Maintaining it is the rest of the hole.

When a founder has actually spent their career in the trenches of instructional design, they build for a different set of problems. They aren't just thinking about how to get the content onto the page faster. They're thinking about what happens next.

What happens when the source material changes? What happens when a new piece of legislation drops, or a clinical procedure is updated? The real killer isn't the cost of creating training; it's the cost of keeping it alive and accurate.

That's why when we started building CourseAgent, the obsession wasn't simply 'brief-to-course'. It was 'compare-and-update'. What if the software could read the old policy document and the new one, and highlight precisely which five paragraphs in your twenty-section course are now wrong? What if it could suggest the updated wording, in the right tone of voice, and let you accept or reject the changes in a click? That’s not a faster way to hit from the tee. That’s a tool for playing the second shot.

It’s a less glamorous problem to solve. It’s harder to explain in a thirty-second clip. But it’s the one that actually keeps L&D professionals awake at night.

From words on a screen to learning that sticks

There’s another level to this. Most creation tools are good at one thing: generating text. They are, essentially, very fast, slightly uncritical interns. They can fill a PowerPoint slide with bullet points faster than any human.

But learning isn't just the transmission of information. If it were, we could just email everyone the textbook and be done with it. Learning is about structure, challenge, and application. It’s the difference between reading about a golf swing and feeling the change in impact when you keep your head down.

When a practitioner builds a tool, they bake the craft into it. They know that a course needs more than text and a multiple-choice quiz. It needs scenarios that force decisions. It needs activities that make the learner classify information, not just recall it. It needs to test application, not just memory. You build in the principles of good design - things like Bloom’s Taxonomy, not as a theoretical checklist, but as the underpinning of the interactive components you make available.

You build in the ability to create dialogues where a learner has to choose the right thing to say to a distressed patient or an unhappy customer. You build a hotspot reveal that lets an engineer learn the parts of a new machine. You build a timeline that shows how a policy has evolved.

These aren't decorative features. They are the mechanics of learning. They are the things that turn words on a screen into a tangible change in someone’s skill or understanding.

An AI tool that only generates text is the equivalent of a caddie who just hands you a driver on every tee, regardless of the hole. A tool built by a designer asks what the shot requires. It asks if you need to shape it left-to-right. It asks if laying up short of the water is the smarter play.

It respects the player, and it understands the course.