How AI Will Affect Instructional Design: My Thoughts Right Now
Early thoughts on how AI will affect instructional design. Most ID jobs won’t disappear, but the field will change as AI augments our work.
Right now, it’s impossible to truly know how AI will affect instructional design in the future. The reality is that this technology is changing too quickly, and no one really knows what will happen. I’ve definitely heard from folks that are scared the whole field of instructional design could disappear and we’ll all be replaced by AI. I’ve also seen predictions from people optimistic about AI who expect it to transform and improve the whole field within two or three years.
While I’m loathe to make too many predictions, I expect the reality will be somewhere in the middle. I don’t think most instructional design jobs will disappear, but I do think the field will change as we integrate more AI tools to augment our work.
AI to augment and adapt, not replace instructional designers
For the most part, AI tools won’t replace instructional designers. One of the challenging aspects of this field is that it covers a lot of different tasks and requires a wide range of skills. We need to do analysis, writing, working with SMEs, visual design, etc. What makes instructional design challenging also means it’s harder to replace us.
In general, I think instructional designers will use AI to augment their work, not replace them. Instructional designers who learn to adapt their practices and processes to use AI will likely be more successful.
Self-driving cars and adaptive cruise control
Think about technology in cars. Self-driving cars have been promised as “just a few years away” for more than a few years now. But driving is a complex task that requires a bunch of different processes and reactions. While there are some cars that can operate independently in some conditions, the technology still has a lot of problems. We’ll continue to see research and advancement in the field of self-driving cars, but the safety and other concerns mean it won’t be widespread for a long time.
However, while we don’t have fully self-driving cars, we do have driving technology that is helpful. Take adaptive cruise control, for example. Instead of setting a single speed on your cruise control and manually adjusting when you need to slow down, adaptive cruise control adjusts the speed for you. You set the maximum speed, and then the car adjusts based on the distance between you and the car ahead of you. The technology takes an annoying task that used to require manual effort and makes it easier.
In the short term, the improvements in AI for instructional design are going to be more like adaptive cruise control than self-driving cars. AI tools can help you with certain parts of your role, but they won’t replace you. It’s worth finding the ways that AI can make your job easier.
AI voices
While instructional design overall isn’t going to disappear, the voice over field is going to be hit hard. AI voices are much better now than they were even a year ago, and I continue to see new tools and improvements all the time. I know several IDs who are already using primarily or exclusively AI voices.
Personally, I think the AI voices are OK for short content, especially didactic or instructional narration. They still fall short for scenarios and more emotional or conversational content though.
For example, I recently had a script for a Vyond video where one character exclaims, “Typical!” In my voice over script, I included a direction for the voice over artist to roll her eyes while she said the line. And yes, I could “hear” that body language in the final recording. We are a ways away from being able to get that kind of nuance from AI voices. It was absolutely worth hiring a voice actor to humanize that script.
I expect that elearning will continue to shift to AI voices because they’re “good enough” for a lot of content. I don’t think real voice over is going to disappear entirely, but I think it will become more of a premium product.
Tech writing and software training
While I don’t think ID is going to go away, I think jobs designing software training will be affected. Tech writing, software training, and procedure documentation jobs won’t disappear completely, but AI can perform really well in these areas.
A large percentage of technical documentation and software training will be initially captured with tools like Scribe AI and guidde. People will be needed to identify what’s needed, record the process, and clean up the results. But even right now, the tools can speed up the process enough that one person can do work that used to take a whole team.
Faster content generation. Should we?
In a recent discussion on LinkedIn, someone talked about how we can now generate content in minutes with AI. We can…but do we want to?
Faster creation of lower quality content isn’t a win for the L&D field. Some organizations will get seduced by that. It’s just like how some organizations have opted to offshore work and then been disappointed in the results.
When you see how fast AI can create content, it’s initially amazing…until you look at it in detail and start to see the flaws. Sometimes it would take more time to edit and improve AI-generated content than to just write it yourself.
In practice, I think people will find ways to be more efficient in certain parts of the process of creating training. However, most companies won’t find AI-generated training valuable enough to pay for. Yes, you can generate content quickly with AI. You still need to spend a lot of time reviewing and correcting it. SME review becomes even more important for AI-generated content because it’s so critical to have someone check the accuracy.
Improving efficiency in creating training
I am finding it helpful in certain areas, and AI does improve efficiency in some tasks. For example, Claude is really good at taking big storyboards or scripts and turning them into nicely formatted voice over scripts with file names. That’s a boring administrative task that I used to spend time doing manually. Now, I can give Claude a prompt and a document, and then work on something else while it sets up a clean script for me.
Robyn Defelice is working on updating the ATD research on how long it takes to develop training. I’ll be interested to see what Robyn’s newest research finds. My prediction is that AI won’t have shifted things much yet in this round of the research. Most organizations are just dipping their toes into AI. Certain areas will go faster (like AI voices), but most of the rest is currently experimentation and piloting, not production. I don’t think the use of AI is widespread enough to really shift the time estimate benchmarks yet. (But we’ll see when Robyn publishes the updated version–maybe I’ll be wrong!)
Adaptive learning experiences
Another way AI will be used in instructional design is creating adaptive learning experiences. In the short term, I expect this to look like some of the AI-enabled elearning Garima Gupta has been demonstrating. Many of us in the field have wished for years that we could automatically assess and provide feedback on open-ended responses in elearning. Now, it’s possible to create those exercises, giving the AI tool a rubric or standards to use to provide feedback.
Chatbots are another way to provide adaptive learning experiences. Character.ai is careful to remind you that everything it does is made up, but you can use that right now to get personalized language learning, a virtual partner to practice interviewing, or help with writing. Especially in parts of the world where teachers are in short supply, this kind of personalized learning experience could make a huge difference for learners.
Right now, most organizations seem to be approaching AI-generated content with caution, and I don’t think they’re wrong. One of the ways AI will affect instructional design will be that we will curate and create these chatbots and adaptive experiences. AI tools will need training and guardrails before they’re ready for widespread use in organizations, and that’s a role instructional designers can potentially do.
How AI will affect instructional design: Short-term and long-term effects
The general trend with new technology and predictions is that people overestimate the effect of technology in the short term and underestimate the effect in the long term. I think that’s a lot of what we’re seeing now in how people assume AI will affect instructional design. People assume we’re going to see huge changes right away, but it’s actually going to take longer.
Given that, we’ll probably see some small improvements in efficiency in the process of creating training in the next few years, with larger changes in a few specific areas like AI voices. I also think we’ll start to see AI used in learning experiences like chatbots and personalized learning, but most organizations will move cautiously.
The big revolutionary changes won’t really be visible for 15-20 years. I don’t think anyone can accurately predict that now. AI feels like a big shift, but it’s too early to really be able to see what things will look like that far out when we’re so early in this process of change.
I’m putting these early thoughts on my blog now with the intention of revisiting this down the road. Maybe a year from now I’ll realize I was wrong, and I completely underestimated how fast everything would shift. But based on everything I’ve seen so far, I don’t think the changes will be drastic yet.
Upcoming events
Morning Buzz: Designing and Developing Scenarios. Tuesday, April 23, 7:30-8:15 EDT. I will be facilitating an open-ended discussion about designing and developing scenarios at the Learning & HR Tech Solutions Conference.
Generating Plausible Choices and Consequences for Scenarios Using AI Tools. Thursday, April 25, 10:00 EDT. This is a hands-on session to learn to use AI tools to generate draft scenario questions, choices, and consequences. Practice refining prompts, recognize the limitations of AI tools, and know when to rely on AI versus manual content creation. Part of the Learning & HR Tech Solutions Conference, April 23-25 in Orlando. Save $500 your registration with my exclusive promo code SPEAKER129.
Christy, I am really disappointed to see that line about how some companies outsourced ID work only to regret it later. I strongly believe there are good and terrible IDs everywhere. I am really not sure what being in a specific country has to do with a skill. In India, we learnt from the best IDs in the West about 30 odd years ago and then got quite good at it. Indian companies routinely win international awards. Anyway, you’re entitled to your opinion. I am just sorry I found out about it.
You’re right, I could have written that sentence with some more nuance; saying “some companies” wasn’t clear enough to show that it’s not a universal experience. There are exceptions to the general rule about outsourcing ID and elearning development work. I worked with a good elearning development company in the Ukraine a number of years ago, and that was a positive experience. Upside Learning in India seems to be doing good work as well.
But often, companies accept lower quality work from offshore vendors, and they view it as the tradeoff for lower costs. The good examples are the exception to the general expectation based on overall experiences.
For example, when I worked at Accenture, my job was to QA the work done by Indian IDs. We knew the Indian IDs wouldn’t meet the standard of American development, and so we planned time for me to review it, identify all the mistakes, and get them fixed. Some IDs were better than others, but it wasn’t uncommon for modules to have dozens and dozens of errors in them. This was very basic, straightforward procedural software training, not something complex. My manager knew there would be extensive problems, but it was cheaper for her to hire IDs in India to create the first draft, and then to have me do the QA. We planned to spend time to get the quality up to an acceptable level.
I’ve worked with multiple clients over the years who have hired LMS vendors in India. Over and over, clients tell me that they selected those vendors because the price looked better on paper. However, while the hourly rate was lower, it took them three times longer to do anything, and they weren’t responsive to problems. This isn’t with just a single vendor; it seems to be the current state of the Indian LMS vendor market.
The specific incident I had in mind when I wrote that sentence isn’t either of the above examples though. A few years ago, a prospective client contacted me to discuss a vision for a large training program. She had some great ideas, and I thought the overall plan sounded like something where I could really provide value. Unfortunately, she rejected my proposal for the first phase of work because it was too expensive. She worked at a global company and was used to working with people in many countries, so she found vendors in India and the Philippines to create the elearning instead. That’s fine with me–I tell prospects that if they can meet their business goals with a cheaper vendor, then they should do so. I’m not the right person for every job.
A year or so later, she reached out to me again. She’d been round and round with multiple offshore vendors. None of them could provide the quality she wanted, and they all required a huge time investment from her to project manage them and keep them on track. She genuinely tried to make it work with those vendors, providing feedback and examples of what she wanted. Ultimately, the quality she wanted was beyond what they were capable of producing. So, she grudgingly came back to me, knowing that she’d have to pay more to get it done right.
Maybe it’s the selection bias of my position, where people come to me when things aren’t working with other vendors. But I only hear stories of when it goes wrong; I never hear anyone talk about how they’re delighted with the quality of the elearning they get from Indian vendors. I hear that it’s “good enough” for the price sometimes, but that’s about the most positive anyone says.
I know there are folks trying to improve the field in India. It sounds like you’re probably in that category of people. If so, I wish you well. I think the L&D field as a whole benefits when there’s better quality competition.
I found this blog post to exceptionally interesting and informative. I am pursuing my degree in Instructional Design (ID) and to be honest I hadn’t thought much about how AI would or could affect my future career. I have previous teaching experience, and I would have to agree with you that AI voice overs would lull a lot of people to sleep because they do not have that emotional capacity that you talked about. There is something to pitch, tone and enthusiasm that can capture the hearts and minds of the individuals engaging with a product. You also mention that creating training developments cannot be solely entrusted to AI because it is a large collection of work that will need to be reviewed and edited for accuracy by ID and SMEs. As IDs, I am assuming it would be our duty to ultimately ensure that the learning objectives are specific, measurable, achievable, realistic and timely all the while ensuring that the needs and wants of the stakeholders are being met. It sounds like AI is just not equipped to handle this with 100% accuracy…yet. Stakeholders seem to be risking the quality and timeliness of the product development if they rely solely on AI. Your points ensure me that I am still making the right choice in pursuing my ID degree and that I will still be relevant in this industry in the next 10 to 20 years. The exciting things to look forward to are the chatbots for learning experiences. In essence, chatbots provide an avenue for role play and can be useful in practicing human interactions or procedures in the context of the desired skill. Overall, I found this topic to be incredibly interesting and I look forward to seeing the advancements in AI as the years move forward. Thanks again.
Thanks for commenting, Elizabeth! This will probably be addressed in your graduate program, but we don’t generally use the SMART format for learning objectives. That’s a common format for goals, but for learning objectives you’ll probably learn the ABCD format (Audience, Behavior, Conditions, Degree).
Even right now, with no improvements in technology, ChatGPT can be taught to write objectives in that format. Karl Kapp shows this in one of his recent videos on AI: https://youtu.be/T6aEbmiXWW8?si=LTSOd9zImbiM43Mn&t=272
But, even writing it in that format doesn’t mean it’s the right objectives. Generative AI tools don’t understand the context, and part of our role as IDs will be making sure that the AI content is relevant and realistic.
Most organizations seem to be pretty cautious about AI now, other than the vendors creating tools that use or incorporate AI. I think most stakeholders are cognizant of the risks and aren’t jumping in completely with AI-generated content just yet.
I am so grateful to you for a thoughtful, informed (while asserting you are not an expert) overview of this topic. I am encouraged to step out of my comfort zone to try a few more AI tools carefully. I am less inclined to be bullied by giddy promoters exhorting hasty adoption because all the experienced and respected Instructional Designers advise me to proceed with educated caution regarding AI.
Most of all, this blog is a huge help to me right now as leadership in my workplace is eyeing my process for speed. They have heard from AI marketers that training content can be “easily generated by anyone” using the latest fabulous AI product. Our customers can be demanding with deadlines, but many are utterly unforgiving if their time is wasted by misinformation. We could lose customers if I were to let an untrained, excitable puppy on steroids run amok with technical content. I have been advised by AI developers contributing to the open source tools that the LLM and ML is moving faster than they would prefer.
Thank you for writing this blog.
Thanks, Julia! There are folks like Almira Roldan and Donald Clark who have been talking about AI for years before the most recent push. Those are the people I view as truly experts. I see too many folks on social media claiming to be instant experts, and I want to separate myself from those voices. But, I think it’s still useful to speak as a practitioner who is experimenting tools and watching the trends.
The marketing hype for AI is a LOT right now, and you’re smart to be cautious. I’ve sent some pretty blunt messages to marketers who have contacted me with wild claims. I love your metaphor of AI as “an untrained, excitable puppy on steroids run amok with technical content” because that is such a vivid image to convey the risk. We already have problems being trusted with our training because everyone has had to sit through irrelevant training at some point. AI could amplify that problem rather than solving other challenges.
But, if we can figure out how to use AI to adapt the learning, then maybe we can make it more relevant. Even just letting people skip the content they already know and helping them focus on their gaps would be a huge improvement. People could spend less total time training for the same results–or maybe even the same time training with improved results because of AI-driven practice. That’s an exciting possibility that I’m looking forward to exploring more.
Coming from an instructional designer, I’m surprised there are so many unfinished editing issues with this article before it was published.
There are significant misspellings and typos.
As an instructional designer, I’m sure you understand that different mediums and purposes have different writing styles and needs. As a blog by an individual person (not an organization), my writing here is always pretty informal and rougher than it would be for more polished publications. My blog is my first draft of thoughts. When I write for publications or training, I spend more time editing than on the blog content I put out every week. Most professionals in the field understand the difference in platforms.
But, since you commented, I ran this post through Grammarly to double check for typos and errors.
aanalysis s/b analysis
Misplaced hyphen fully self-driving
year s/b years
to revisit, Grammarly recommends “of revisiting” (which I agree flows better)
That’s four errors total in a 1700 word post, which is maybe a little higher than my goal, but not unreasonable for first draft writing. The rest are all style choices (e.g., elearning rather than e-learning), not errors. It might not match your organization’s style guide, but I don’t work in the same place you do.
To be honest, I don’t think any of those were “significant” errors. There was one misspelled word, and it was a clear typo. None of it interfered with how people understood the content. In fact, the overall response to this post has been positive. I’ve received quite a few substantive replies and shares on LinkedIn, where most blog conversations happen these days.
Do you have anything to say about the substance of the post, or is the one misspelling really the only thing you got out of this writing?
How ridiculous! 4 errors/typos. Please! I didn’t notice any glaring errors (and I’m an ESL teacher and tech writer), and if they were there, they certainly didn’t affect the quality of the information and its substance. Besides, it’s in more of the informal blog style which is far more engaging.
Thanks, Nick, I appreciate the support. Casey’s comment caught me by surprise. In 15+ years of blogging, I’ve made plenty of mistakes. This post didn’t have anything glaring though, and certainly not anything that warranted a comment like that.