AI in Course Creation, Honestly

I sat through a demo for an AI course creation tool the other week. Not ours. A competitor’s.

The presenter dropped a 50-page PDF into the system, and in about ninety seconds, it spat out a ten-module course, complete with introductions, summaries, and a twenty-question final assessment. The virtual room was audibly impressed. It was, on the surface, an act of magic.

But as I looked at the output, I felt that familiar, quiet sinking feeling. The feeling I get when I see something that is technically impressive but misses the point entirely. The content was accurate, the grammar was perfect. But it was inert. Lifeless. It was words on a screen, information arranged in order. It wasn't teaching.

A scenario question about handling a customer complaint was so generic it could have applied to a bank, a bakery, or a builder's merchant. It lacked the specific, gritty reality that makes learning stick. It tested whether you'd read the previous paragraph, not whether you could handle a furious human being on the other end of the phone when a delivery hasn't arrived. There was no craft.

The experience crystallised something for me. The conversation around AI in course creation has become deafening, and most of it is about speed. About the magical, instantaneous conversion of documents into 'courses'. But we're asking the wrong questions. We're so dazzled by the speed of production that we're forgetting to ask about the quality of the learning.

So, from a practitioner’s standpoint, stripping away the marketing noise, what is AI genuinely good at? And where are its limits? Honesty is required.

Where the Machine Actually Helps

For a professional who knows what good looks like, AI can be a powerful lever. It’s an assistant – tireless, absurdly fast, but one that needs clear direction. I have found its utility falls into five distinct areas.

1. Defeating the Blank Page Every writer, every designer, knows the tyranny of the empty screen. AI is exceptionally good at providing that first, rough draft. Dropping in a pile of source material – policy documents, slide decks, technical manuals – and getting back a structured outline in moments is a genuine productivity leap. It’s not the finished product. Not even close. But it’s a starting point. It gives you something to react to, to shape, to disagree with. It turns a creation task into an editing task, and for many, that’s a much lower barrier.

2. Automating the Grunt Work There is a huge amount of low-value, repetitive work in course design. Writing multiple-choice questions is a classic example. An AI can generate twenty variations on a theme faster than you can drink your coffee. Are they perfect? No. You still need a human eye to check for ambiguity, to ensure the distractors are plausible, and to confirm they test the right thing. But it gets you 80% of the way there. The same goes for creating summaries, glossaries, or rephrasing a complex sentence for a different audience. It’s an accelerant for the tasks that don’t require deep pedagogical judgement.

3. Consistency at Scale When multiple authors work on a large curriculum, 'voice drift' is inevitable. Terminology shifts, tone varies, formatting goes astray. A well-instructed AI can be a powerful enforcer of consistency. It can rewrite sections to adhere to a specific style guide, ensure key terms are used correctly every time, and apply formatting with inhuman precision. In a 200-module curriculum for a financial services firm, this isn't a trivial benefit; it’s a core component of quality assurance.

4. Translation and Localisation This, for me, is one of the clearest and most transformative wins. Historically, translating a course was a significant project. It was slow, expensive, and often limited a company’s global reach. AI has fundamentally changed this. The ability to take a course and render it in nineteen different languages, from the text to the interface buttons to the quiz feedback, in a matter of minutes, is profound. It democratises access. Suddenly, a vital piece of health and safety training developed in the UK can be used, with very little friction, by teams in Poland, Brazil, and Japan. This is not a small thing.

5. The Living Course: Compare and Update This is the big one. The one that shifts the paradigm. Most organisations see courses as static objects. You commission one, you deploy it, and it slowly decays as the world changes. The biggest problem in corporate learning isn’t creating the course; it’s keeping it current.

This is where AI excels. It can take a new piece of regulation – say, an update to the FCA’s Consumer Duty or a new NHS directive – and compare it against an existing course, flagging precisely which sections are now out of date, and even suggesting the updated text. This transforms a course from a disposable, depreciating asset into a living one. It’s the difference between a multi-week, £15k rewrite project and an afternoon’s work. It’s why we centred CourseAgent’s design on this capability. It solves a structural problem, not a superficial one.

Where the Human is Irreplaceable

For all its power, AI is a tool. And like any tool, it has its limits. Believing the hype that it can replace the designer is not just naive; it’s dangerous.

1. Real-World Context and Nuance I return to that generic customer complaint scenario. AI doesn’t know your company culture. It doesn’t know that 'Dave from accounts' is the person to call, not the 'finance department'. It has never sat in your canteen and heard how people really talk. It can't create a scenario that feels authentic to your people because it has no experience of their world. The most powerful learning often comes from scenarios that hold a mirror up to the organisation's specific reality. That requires a human who has listened, observed, and understood.

2. Genuine Pedagogical Judgement You can tell an AI to structure a course according to Bloom’s Taxonomy. It will dutifully label sections with 'Analyse', 'Evaluate', and 'Create'. But it doesn't understand what those words mean in practice. It cannot make the crucial judgement call that a particular cohort of learners will struggle with a concept unless it’s introduced via a specific metaphor. It can't look at an assessment and say, “Technically this is correct, but it doesn't actually test if a new manager can have a difficult conversation, it just tests if they remember the five-step model from the PDF.” That is the craft of the instructional designer, earned over years. It is an act of empathy and intellectual translation, not of computation.

3. Accountability If an AI hallucinates and includes a critical error in a compliance course – what happens when an employee follows that incorrect advice and your organisation is fined? Who carries the can? The answer is, and always will be, the organisation and the human professional who signed off on the training. The AI has no professional indemnity insurance. It has no duty of care. It is a system for generating probabilities in the form of language. The final, accountable sign-off on whether a course is fit for purpose is a human responsibility. Abdicating that responsibility to a machine is a catastrophic failure of governance.

AI is not a replacement for expertise. It is a powerful, flawed, and brilliant amplifier of it. The best course creators won't be replaced by AI; they will be the ones who learn to master it, using it to automate the drudgery so they can spend more of their time on the things that matter: context, empathy, and judgement.

It allows the craft to be focused where it counts.