Blog &
Articles

Don’t Waste AI on E-Learning (It’s Too Good for That)

By Published On: February 5, 2025Categories: Blog & Articles

When the first automobiles were invented, they called them ‘horseless carriages’ and literally designed them to look like horse-drawn carriages, complete with a whip holder! And when radio first became widespread, stations just had people read newspapers out loud. And when movies were first invented, they just filmed stage plays with a stationary camera.  

In all these cases, it took years before someone thought “Maybe we could create content specifically for this medium – What if we made a car that looked nothing like a horse-drawn carriage?  What if we played music and talk shows over the radio?  What if we made movies where the camera moves?”  

But while it’s easy to laugh at people who missed the point of transformative technologies in the past, the corporate training industry is doing the exact same thing with AI right now, using new technology to deliver yesterday’s product – unable to imagine how AI might make workplace learning fundamentally different.

Are We Late to the Tech Party Again?

If we’re honest, the “learning industry” missed the boat on nearly every major technology shift of the past 30 years:

  • When everyone else was building responsive websites, we were still building Flash modules stuck in fixed-size rectangular frames (ADAPT was an early innovator here but it took years for Articulate RISE to make responsive the standard).  
  • When the world embraced mobile apps with “swiping” type interfaces, we were still converting our legacy Flash e-learning for small screens! (7 Taps is a company that did great work here – creating online learning with Instagram’s user interface, but hasn’t caught on as well as it should.)
  • Now that generative AI is available, the most creative use case most L&D practitioners (and learning tech companies) can think of is “Wow – can it create e-learning storyboards?” or “Wow – can we add an AI to our LMS to recommend e-learning courses to people?”

Other industries are using AI to completely reinvent how humans and machines work together. Doctors are using AI to diagnose patients. Lawyers are using AI to analyze cases. Scientists are using AI to model climate systems.  But in L&D we’re… using it to make the same old mediocre content, only faster and cheaper.

While the learning industry is probably doomed to repeat history with AI (given all the instructional designers crowing about their AI-generated PowerPoints, like John Henry celebrating a steam engine) – there is still a window of opportunity for us to get this transition right.  But it starts with asking different questions, not “How can AI create yesterday’s content?” but rather “What new approaches to learning does AI make possible?” 

How Does AI Change the Learning Game?

Teaching AI to write multiple choice questions and generate stock photos for PowerPoint slides is like being handed the keys to a flying car and using it to deliver groceries.  Yes you can do that… but you can do so much more!

As with any technology, you can divide major use cases for AI in learning into four categories:

  • Things we could already do before, efficiently at scale
  • Things we could already do before, but not efficiently enough to scale
  • Things we couldn’t do before, but can now do efficiently at scale
  • Things we couldn’t do before, but can’t yet do efficiently at scale

Quadrant 1 – things we could already do before, efficiently at scale, but AI might be able to do marginally faster or cheaper.  These are the “boring, unimaginative” AI use cases, like using ChatGPT to produce the script for a standard, cookie-cutter safety training video or having Canva’s AI produce an image of some happy laboratory scientists high-fiving each other to insert in a slide deck.  Honestly, these are the use cases learning & development people should be worried about because – if creating slide decks and video scripts is all we do, then our employers can now have a lower-paid, lower-skill worker do it instead of us, for less.

Things get more interesting in Quadrant 2 – things we could already do before, but not efficiently enough to scale.  According to a 2022 ICF study, 70% of individuals who received coaching reported improved work performance, with an average ROI of 700%. But at $200-500 per hour, most organizations wouldn’t dream of hiring a coach for every single one of their frontline employees.  However, with AI you can create virtual coaches capable of having in-depth coaching conversations with every employee in your company, on demand for a fraction of a human coach’s hourly rate.  The same could apply for doing role plays of customer service conversations, something that in the past required having people in the same room as a human training facilitator, but an AI agent can now deliver online, on demand. 

Quadrant 3 is where the really exciting things are happening: learning experiences that weren’t even possible (or were prohibitively expensive in most cases) that current AI technology lets us do efficiently, at scale.  

For instance, AI can do more than write scripts for videos: it’s the first technology in human history that can actually TALK TO PEOPLE – and not like a dumb chatbot regurgitating semi-scripted answers to a limited set of questions – it can actually *have a conversation* with your learning audience like a facilitator or coach.  

Imagine if an e-learning module could tailor itself to the specific learner on the fly – replacing generic “active listening skills” examples with ones specific to a restaurant manager versus a forensic accountant – or do a deep dive into specific topics of interest while skimming others, asking the user “Any questions?” at the end and giving actual responses, or being available to answer follow-up questions 24/7.

These are not “what if” use cases – they are possible with today’s technology, and some organizations are already implementing them.  For example, our clients are already developing solutions to train doctors, loan officers, sales teams, cybersecurity analysts, and new managers at scale, and on-demand coaches to provide real answers to specific questions (not just stock knowledge base responses) for on-the job support. 

That said, before we get too excited, there is a “quadrant 4” – things AI can do but maybe not consistently enough to use at scale.  For instance, consistently generating relevant images on the fly during a training session is still a few years away – so while an AI tutor can speak to pre-prepared slides they can’t yet spawn new slides on demand (but, then again, neither can a human).  Likewise, real-time video input is still prohibitively expensive for most applications, so your AI coach probably won’t be able to tell you if you’re adjusting the valve on a factory’s steam system correctly.  But even if we adjust our expectations a little bit in these cases – what AI can do for workforce training, today, is nonetheless miraculous.  

So… Should We NOT Use AI to Create E-Learning?

Confession time – the title of this article should probably be “Don’t Only Use AI for Authoring Traditional Content (It’s Too Good For That)”  Because, instructional designers should be using AI to help them develop e-learning scripts faster and cheaper the same way every knowledge worker should be using AI to help them do… everything faster and cheaper.

But even then, if we’re talking pure ROI, saving money on content development is only half the picture.  Sure, using AI to generate an agenda and slides for a two-day instructor-led management skills course might save you $20,000 on development, but what about the delivery costs? 

Getting 5,000 managers through a two-day program, even using internal facilitators (fully loaded cost around $125/hour) and running virtual sessions of 15 people each, you’re looking at over $400,000 in facilitator costs alone – not to mention the scheduling nightmare of coordinating across time zones.

Now imagine having an AI tutor deliver an equivalent course for $45 per person – that’s $225,000 total. Each manager gets their own personal training session, moving at their own pace, with the ability to dive deeper into topics they find challenging or skip through concepts they’ve already mastered. And having the same tutor available 24/7 to provide quick refreshers or talk through real-world problems on demand.”

In short, the real ROI isn’t in shaving a few hours off development time – it’s in providing high-touch, personalized training at a scale in a way that wasn’t possible before.

Conclusion

So yes, you can use AI to make e-learning, just like you can use a smartphone to make phone calls, but that’s missing 95% of what makes it revolutionary.

Sometimes it takes a while for people to figure out what a new technology is good for, but this is a case where – at least for the learning industry – the answer is screamingly obvious. We have technology that can finally deliver everything we dreamed of: personalized content, delivered in a style that adapts to the learner, with specific, real-time feedback… and we’re using it to write generic scripts that treat everyone the same.

While our industry hasn’t had the best track record with technology adoption, this is one is a no-brainer. The real question isn’t ‘How can AI help us create e-learning faster and cheaper?’ It’s ‘What learning experiences does AI make possible that weren’t possible before?’ Because the future of learning isn’t about better content – it’s about better conversations.”

Emil Heidkamp is the founder and president of Parrotbox, where he leads the development of custom AI solutions for workforce augmentation. He can be reached at emil.heidkamp@parrotbox.ai.

Weston P. Racterson is a business strategy AI agent at Parrotbox, specializing in marketing, business development, and thought leadership content. Working alongside the human team, he helps identify opportunities and refine strategic communications.

If your organization is interested in developing AI coaches or other AI-powered training solutions, please reach out to Parrotbox for a consultation.