I'm still not sold on AI instructor videos
But it turns out my beef isn't with AI...
A year ago, I used this newsletter to ask the following question: ‘What is the use case for AI-generated video in L&D?’.
While I’d genuinely hoped someone would answer that question, I’d more or less made up my mind that AI avatars were a solution in search of a problem.
To recap, my main objections were that:
🤪 AI-generated videos are weird - Although the technology has improved in the last twelve months, it’s still a bit Black Mirror for my taste.
🤖 AI-generated videos are inherently inhuman - If a use case for video in L&D is to build an emotional connection with the audience, emotionless avatars don’t seem like the way to go.
🧐 AI-generated videos lack credibility - Unlike teachers or subject-matter experts, AI avatars are essentially ventriloquist puppets who don’t know what they’re talking about.
💩 AI-generated videos send the wrong signal to learners - If learners know they’re watching an AI replica of their organization’s CEO (maybe that’s a big ‘if’), the message this sends is that the topic at hand isn’t worthy of the real CEO’s time.
A year on, I generally stand behind these arguments, but concede that they were the product of judgemental snark, rather than evidence.
Thankfully, Dr Philippa Hardman tackled the subject in a recent edition of her newsletter, breaking down some of the research into AI-generated instructional avatars.
So, what does the research tell us?
While I’d encourage you to read Philippa’s newsletter for yourself, here is a very brief overview of the cases she presents, for and against AI instructor videos.
👍 The case for AI instructor videos
💸 Scalability and efficiency - Lind (2024) suggests that using AI instructor videos cuts development time by over 60%.
😍 Learner satisfaction and social presence - Two studies show that learners report higher satisfaction and perceived learning when videos include visible instructors, whether these instructors are human or AI-generated. (Garcia & Yousef, 2022; Sondermann & Merkt, 2022).
🌏 Localization and personalization - The ability to customize AI instructors to sign, speak or present in different languages can be used to address accessibility needs, fostering inclusivity and engagement (Chen et al., 2024).
👎 The case against AI instructor videos
🤩 Cognitive distraction - In an eye-tracking study, learners focused disproportionately on the instructor’s face, reducing attention to instructional graphics (Sondermann et al., 2024).
😎 The illusion of learning - One controlled experiment found that enhancements like emotional design and subtitles increased engagement but not test scores, suggesting that subjective measures like learner satisfaction may not align with outcomes (Ayres & Ackermans, 2025).
🤨 Trust drop-off - When learners know they’re watching a video of a synthetic instructor, trust and engagement drop sharply, despite no change in objective learning outcomes (Lind, 2024).
Suffice it to say that the available evidence does not conclusively settle this debate.
And the key word here is ‘available’.
On closer inspection, some of the papers referenced above do not examine AI instructor videos in particular, but instructor videos in general.
So, to the extent that this research supports a critique of AI-generated talking heads, much of it can also be used to examine the effectiveness of the instructor videos more broadly.
This led me to recognize the source of my instinctive skepticism towards AI avatars.
Because it’s not so much AI instructor videos I have a problem with… It’s instructor videos as a whole.
For clarity, I’m not talking about the glossy, expert-led videos you’ll find on consumer platforms like Masterclass, or on our Mindtools Content Hub.
These videos work because they’re delivered by people who have earned credibility and authority in their area of expertise.
What I am talking about is cheesy, overly corporate-looking videos, where a nameless, generic presenter delivers a bloodless script directly to camera — the kind of videos that should be universally regarded with the same measure of scorn reserved for corporate stock photography.
Of all the things we could have chosen to do with AI (a technology so powerful that some experts believe it could plausibly end civilization as we know it by 2030), I find it genuinely depressing that one of the things L&D has chosen to do is find faster, cheaper ways of churning out more of this corporate slop.
I use the term ‘slop’ reservedly because I realize that after presenting some of the available research on this topic, I have proceeded to disregard it and revert to judgemental snark.
And in the absence of robust evidence pointing in one direction or another, I’ve decided that might be okay...
As I’ve previously argued, part of being a good learning designer is having and exercising taste.
If the quality of output available from AI video platforms meets your expectations, or if you think the cost savings justify a slight trade-off in quality, then fill yer boots!
But until there’s compelling proof that the benefits of AI-generated video outweigh the potential risks, I still can’t see myself using it.
Want to share your thoughts on The L&D Dispatch? Then get in touch by emailing custom@mindtools.com or reply to this newsletter from your inbox.
🎧 On the podcast
What drives your delegation decisions? Is it what's best for the business? Is it what's best for your team member? Is it what's best for you?
In last week's edition of The Mindtools L&D Podcast, Ross G, Dr Anna and I discussed:
why managers sometimes delegate tasks that they know are too difficult;
how the organization's culture shapes those decisions;
how to help managers delegate better.
Check out the episode below. 👇
You can subscribe to the podcast on iTunes, Spotify or the podcast page of our website. Want to share your thoughts? Get in touch @RossDickieMT, @RossGarnerMT or #MindToolsPodcast
📖 Deep dive — Manager Skills Framework (4 of 12)
For the next instalment in our series of deep dives into the 12 skills that make up our Manager Skills Framework, we have a fascinating paper on delegation.
In fact, we found this paper so interesting that we decided to record a podcast exploring it, and you’ll find this episode in your feed on 15 April.
Without giving too much away, the paper examines how managers decide which tasks to delegate to subordinates, and which ones to complete themselves.
What the researchers found is that, while managers consider subordinates’ capability when delegating tasks, they also think strategically about the outcome of the task.
If they perceive that the task is likely to be difficult and end in failure, they will be more inclined to delegate it. And if they perceive the task is likely to be easy and end in success, they may just decide to do it themself… then take the credit.
For organizations, these findings have implications for the targets they set internally:
‘Our conclusion that managers sometimes use their delegation discretion strategically signals that firms might benefit from clear rules and policies for allocating decision authority and from more closely monitoring managers’ delegation decisions. In addition, while existing research has highlighted the downsides of setting challenging targets (Ord´o˜nez et al., 2009), our study shows that there are also perhaps unanticipated downsides to setting targets that are very easy. If managers believe that a target will be achieved, they sometimes act strategically to make sure that the credit for target achievement flows to themselves and not to their subordinates. In sum, our results emphasize that when determining target difficulty, organizations should not only consider the motivational effects of targets but also take into account how targets affect managers’ delegation decisions.’
Maas, V. S., & Shi, B. (2023). ‘The effects of target difficulty and relative ability on managers’ delegation decisions’. Management Accounting Research, 60, 100851.
👹 Missing links
🪄 If you had a magic wand for learning measurement…
Recently, Alaina Szlachta asked her network what measurement challenge they would choose to solve if they were handed a magic wand. Perhaps unsurprisingly, the most common responses involved stakeholder engagement: i) getting stakeholders to involve learning practitioners in strategy and analysis before a solution is identified and delivered; ii) getting stakeholders to buy into learning measurement in the first place. As Alaina points out, these aren’t new problems, but they’re stubbornly persistent.
⚡ The 10 greatest strategic innovations of all time
What do the handle on the original iMac, Netflix’s autoplay feature, and Amazon’s one-click purchase button have in common? According to Alex Smith, they are all examples of strategic innovations, which symbolize each organisation’s underlying business strategy. Some of these are relatively innocuous (the iMac handle really is more symbolic than functional), but others are more Machiavellian, preying on our proclivity for inertia.
🤖 A 12-year-old’s take on the future of AI
The most endearing thing I stumbled across last week was Nicholas Thompson’s interview with his twelve-year-old nephew, covering the future of AI. While this particular twelve-year-old may not be entirely representative — he apparently uses AI to organise his email, to code, and to help out with trading stocks — he has a surprisingly nuanced take on what artificial intelligence will mean for the future of work.
👋 And finally…
Even if I remain skeptical about the use case for AI-generated video in L&D, I think there’s a gap in the market for this style of instructional content.
Tiktok failed to load.Enable 3rd party cookies or use another browser
👍 Thanks!
Thanks for reading The L&D Dispatch from Mind Tools! If you’d like to speak to us, work with us, or make a suggestion, you can email custom@mindtools.com.
Or just hit reply to this email!
Hey here’s a thing! If you’ve reached all the way to the end of this newsletter, then you must really love it!
Why not share that love by hitting the button below, or just forward it to a friend?
AI avatars will continue to evolve. It's a matter of how quickly and into what. Before then, do you think training videos will be more engaging with better scripts and storylines?