The Getting Started Guide to AI for L&D Teams

How AI works. What it took to get here. Whether it's a trend. And how to use it for learning & development.

This is a summary version of this week's article in Yen's Newsletter. You can read the full article here.

Unless you’ve been living peacefully in a cabin with no internet, you’ve probably noticed the hype around AI and the Renaissance it’s creating for talent development leaders. Every L&D person and their dad is talking about AI. Every L&D conference has an AI keynote. Every LinkedIn influencer has new ChatGPT prompts to try.

It can feel like everyone’s an overnight AI expert. It’s almost sacrilegious to not have an opinion yet or not understand how ChatGPT manages to write perfect haikus. There’s so much information overload that it’s hard to imagine where to start, and how you’ll make that information relevant to everyday L&D tasks. With all the AI excitement and innovation happening on the daily, it’s easy to feel left behind and scared to ask “dumb questions” even if your fellow learning leaders love having the answers.

I want to hit the pause button and say, that’s okay!! I’ve seen dozens of articles on how to optimize your workflow with AI overnight, but less articles that cover the basics of what you need to know about AI and how that specifically applies to learning and development.

So here’s my very brief crash course on AI for L&D leaders and getting started guide for applying AI to your everyday work. If you’re passionate about learning design but slightly overwhelmed by AI, this article is for you. There’s so much ground to cover in this realm, this is probably a Part One.

Join our weekly learning newsletter on remote management, L&D, and company culture.

This guide was written by someone who runs an AI HR-tech start-up.

I feel the need to flag this for a few reasons. First, I’m pretty biased. I think optimistically about AI’s impact on our present and future reality of work. I especially think AI can help People Ops teams do more with less and drive more compassionate leadership at scale. If you’re looking for an opinion piece (for which there are many) about how AI is taking over jobs and how it will hurt the L&D industry, you won’t find that here. I might write a piece on AI anxiety and safety later, but that deserves a whole article if not a book. Overall, I see AI as a great enablement tool, especially for L&D, and I’ll get into its use cases in a bit.

Second, I think it’s crucial to know that AI isn’t a new phenomenon but a steady progression of innovation since the late 1950s. We’ve been working on AI solutions for HR teams since 2019. OpenAI and prompt engineering have vastly democratized access to AI, but AI isn’t a new trend despite all the attention its gotten. On that note, I’d also take overnight AI experts with a grain of salt and always diversify the pool of folks you’re listening to.

Third, I can tell you first-hand about shiny new object syndrome. Excitement around new technology can really distract from the problem you’re trying to solve and do your end-users a disservice. I know this because that was the fatal flaw of Kona’s first product direction: AI-generated work-with-me-guides.

It was late 2019 at UCLA, and our co-founding team left the dorms every day extremely excited by IBM Watson. Watson had famously competed against “Jeopardy!” champion, Ken Jennings, and won. It could do a lot of other things well too, beyond trivia. We were particularly interested in how the language model could analyze written text and spit out Big Five personality profiles. We wanted to use AI for remote onboarding and getting to know teammates in Slack.

You’ll notice that Kona no longer does that. This version of the product required a lot of text, and subsequently read your Slack messages to get an accurate personality reading. It was a bit creepy, a bit too early, and definitely didn’t add enough value to users. The lesson here? AI is not a silver bullet for poor problem-solution fit. Learning designers always need to keep their learners top of mind.

Stupid question, how does AI work?

The title was a trick! There are no stupid questions here! This is the briefest lesson I could manage, so I highly recommend you Google these terms and check out this awesome Vox video that includes beautiful visuals like this one:

Step 1 - Training Data. AI is trained through exposure to a vast data set. You can imagine AI sifting through half a million pictures of bananas and not-bananas, to “learn” what a banana looks like. This metaphor falls apart a little when you realize these banana pictures are made of math and that AI is guessing the right answers over time.

For image models like Midjourney or DALL-E 2, this data set might come from the alt-text tags from images on websites. Or those sort of annoying CAPTCHAs you get when a computer is trying to decide if you’re a bot or not. For ChatGPT, the chatbot was trained on a giant dataset of text, books, articles, and webpages.

Step 2 - Latent Space. When we use words like ChatGPT is “thinking” or that computers have “intelligence” it brings up the idea that AI has a mind. This analogy works, somewhat, but it’s not quite the squishy pink brain matter we’re used to. Instead, think of AI’s “brain” like a multidimensional space (not quite 3D, think A LOT more dimensions) where hundreds of variables can help it map out concepts. This latent space is hard for the human brain to imagine, but this Vox video does a tremendous job visualizing this concept.

Step 3 - Prompting. When we use our wildest imaginations to “talk to” deep learning models, it can feel like we’re asking the AI to use its imagination. In reality, our prompt guides the deep learning model to a particular point in that multidimensional, mathematical latent space. With enough training data, eventually there exists an exact location in this space for haikus about John Cena or a picture of Cena at a tea party with Winnie the Pooh.

AI works a bit of magic here called diffusion, translating a mathematical coordinate to the picture or text output you’re looking for. It may iterate on blurry approximations or pick the next appropriate word based on similar long-form text. With a dash of randomness in the process, you get a different output every time. This variation in the output feels quite imaginative!

How did AI get here? Abridged version.

Let’s dive into some AI history. Remember how I mentioned that AI’s been around since the 1950s? It’s a little hard to believe, but the term was actually coined in the summer of 1956, at a Dartmouth conference. Alan Turing wrote about it too, introducing The Turing Test as a method to test whether a computer is capable of thinking like a human being. The concept of the Turing Test is simple––a human should be unable to distinguish a machine’s answers from a human being’s.

Sound familiar?

For over four decades, researchers persevered. There were periods where AI research flourished (1957 to 1974), and “winters” where funding dwindled. Computers didn’t have the power or storage to process information fast enough. Ambitious goals led to disappointment and impatience. For the most part, the public’s understanding of AI was fueled by science fiction authors and pop culture icons: Isaac Asimov’s “I, Robot” to Stanley Kubrick’s “2001: A Space Odyssey” to “Terminator” to “Star Trek”.

Then IBM’s Deep Blue beat Garry Kasparov in chess in 1997. Then IBM Watson wins “Jeopardy!” in 2011 and Google Brain recognizes a picture of a cat in 2012. In 2018, we had self-driving cars. In 2015, OpenAI was founded. In November 2022, ChatGPT completely transformed and democratized large language models to a wide audience.

Read more on this fascinating topic in these resources:

• The Most Amazing Artificial Intelligence Milestones So Far (Forbes)
• The History of Artificial Intelligence (University of Washington)
• Timeline Visual of Artificial Intelligence

Is AI another (silly) L&D trend?

Any experienced talent development leader can tell you that L&D is heavily affected and excited by trends. We can see this in the many waves of trends that have washed over L&D: online learning, learning management systems, micro-learning, learning about learning, learning in the flow of work, virtual reality…

It makes sense to be skeptical about AI and whether it’ll benefit L&D as a whole, or if it’s something we’ll cringe at. (A moment of silence for the AR/VR craze…

Before we talk about AI and whether it’s a trend, I think it’s helpful to think about why the L&D industry gets so excited by new technology:

1. Driving learning is difficult and labor-intensive. You’re essentially changing how people think and how their behaviors at work. That involves a lot of repetition, content creation, and facilitation. It makes sense that speeding this process up or making it 10% easier through automation is really appealing.
2. Learning isn’t always sexy. Unfortunately, trainings aren’t the first item on a manager or teammate’s priority list. L&D leaders have to bend over backwards to advertise their programs and drive engagement. Technology is shiny and exciting. So why not bring the two together?
3. L&D teams are told to do “more with less”. We’re seeing more L&D teams that have been affected by layoffs, teams that lack subject matter expertise, or teams that are simply told “good luck” when asked to educate the company’s entire frontline manager population. The more you can do, the better.

AI checks off all the boxes above. It’s gone viral in the L&D community and for good reason. At the risk of sounding like a broken record, generative AI can truly transform every aspect of L&D: making it less labor-intensive, learner-friendly, cost-effective, and efficient for everyone.

But I think it’s unfair to call AI a “trend”. Looking at the list of other L&D trends, they were all quite industry specific or failed to gather widespread adoption. L&D isn’t the only function losing their minds about AI and its potential. When ChatGPT launched in 2022, it quickly became one of the fastest-growing consumer apps in history. Unfortunately, AR/VR didn’t have that adoption curve.

Generative AI has fundamentally changed the way everyone works. In this way, L&D isn’t simply forcing a tech trend on learners to drive better program engagement. L&D is adapting to this new AI-first world of work. AR/VR isn’t the best comparison to what we’re experiencing today. AI is more like the introduction of the Internet.

So what can you use AI for in L&D?

“We have this awesome new technology, and you’d be stupid not to use it!” This is the sentiment from a lot of folks on LinkedIn when it comes to applying AI for L&D. But there’s less focus on how we can use AI to its full potential. I see a lot of well-meaning L&D experts stopping after the first use case: content generation.

For better or for worse, I blame ChatGPT. On the consumer side, we’ve learned that ChatGPT and DALL-E are great for editing essays, creating reality-defying images, and drafting sick poems. We get the immediate impression that AI is great for generating learning content.

If you’re getting creative, maybe you’ve found an awesome AI tool that helps you magically create subtitles, turn a script into a video, create expert-level voiceovers, or whip up slides within five minutes. Don’t get me wrong, these are all incredible capabilities that cut dozens of hours from the typical learning designer’s day.

But that’s all content generation. When you stop there, you’re really missing out.

AI can tackle far more than L&D leaders think. Our team created this neat infographic to share the use cases we’ve seen so far:

Here’s each section broken down for scale:

• ?? Content Generation. The most time-consuming parts of the job (think voice overs, subtitles, and getting copy just right) are now sped to lightning speed with AI. An L&D team of one can now do the work of many!
• ?? Analyzing Learning Data. The best programs are rooted in quantitative and qualitative research. Before, that meant dozens of call transcripts and surveys, and hours looking for patterns. Gen AI can spot trends super fast.
• ?? Expert Bots. You can add a new performance consultant or facilitation coach to your team in about as much time as it takes to make a sandwich. Cover your talent gaps or offer learners a robot resource.
• ? "Just in Time" Learning. When we talk about bite-sized learning, we're really dreaming of giving folks the exact right resources at the moment of need. AI makes these dreams a reality, offering live skill assessment and feedback.
• ?? Personalized Learning. With Gen AI, courses can become designed for each user's learning journey. Imagine curated, unique courses that address each individuals needs, not just what was convenient to put in the LMS.

These use cases only scratch the tip of the… you get it. Each of these use cases has dozens of creative capabilities to explore, bots to build, and innovative paths for learners. I’d love to hear if any of these use cases sparks ideas for you and your team!

What AI tools should I try for L&D?

I want to preface this by saying that anything I add into this section will seem outdated and misinformed with the rate that AI is moving at right now. There are new apps being added every day, and it’s honestly hard to predict what we’re missing. That being said, there's innovation happening across the board, and it's sparking a new Renaissance of technology enablement for learning leaders.

Here's our big landscape map of the generative AI for L&D. Inspired by Sequoia Capital's Generative AI Market Map:

Testing out new apps in each section will give your L&D team superpowers. Here’s the different categories of apps that we’ve seen so far:

• ?? Coaching and Leadership Support. Ever wish you could sit with every manager all the time to answer key questions, transform their leadership skillset, and guide them through moments that matter? Coaching is transformative, but simultaneously labor-intensive and expensive. AI has made coaching at scale a reality, giving leaders a robust support system.
• ?? Learning Management Systems. Ask any learning leader and they'll probably have strong feelings about their LMS. At times, it can feel like the majority of time goes into wrangling systems to accomplish learning goals, rather than seeing them as partners in the creation process. Traditional LMS is quickly going out of style as AI enables faster course creation, extremely personalized learning tracks, and continuous skills measurement.
• ?? Knowledge Management. Documentation means nothing if no one reads it. The best knowledge management systems hinge on how easily users can find the answers they need. Until now, HRBPs and L&D leaders have had to point folks in the right direction. AI makes self-resourcing easier than ever, turning documentation-hunting into a conversation.
• ?? Performance Reviews. Measuring performance is a crucial process for driving goals and creating systems of accountability. However, the reality of performance management often looks like hours of aggregation and writing reviews. AI-assisted reviews and feedback have transformed this process, so leaders can focus on what's important.
• ?? Content Creation. Last but certainly not least, AI adds more creativity and speed to content generation. Building content used to look like hours adding subtitles and voiceover to video, designing custom graphics and infographics, and laboring over copy. L&D folks can spend more time on what their content will accomplish, and less time on the boring aspects of creating learning content.

If there are any apps or categories that we missed, reply in the email thread and I’ll update the map over time!

Next steps for applying AI to L&D

If you’re ready to take the next step and dive into the world of AI-first L&D, I have a few tips. These are based on what I’ve seen successful forward-thinking L&D leaders do as they apply their passion for AI to create positive impact in their role.

?? You've reached the end of the summary version of this week's article in Yen's newsletter.

You can read the full article on my new Substack! Every article is free and subscribing helps me see your support.

Thanks for reading Yen's Newsletter.

I’m Yen, co-founder at Kona. My goal is to help every manager be a great leader. You might be managing a team yourself, or supporting better managers at your organization. Hopefully, this newsletter helps you look at the ever-changing landscape of leadership in a new way.

Since late 2019, I’ve interviewed 1500+ remote managers, People Ops leaders, and tech executives to learn how they lead teams and design incredible distributed company cultures. While every company’s different, everyone’s trying to answer the same big question: “How do you enable amazing people to do amazing work, while remote?”

That’s the great thing about big questions, they bring people together. Learning is sharing, and I’ve always looked to share everything we know as soon as we learn it. That's the goal for this newsletter: capture and synthesize all of our remote management learnings in a neat and shareable archive.

Every week, I’ll alternate between self-written articles and interviews with industry leaders.

Subscribe on Substack for full-length articles, delivered every Monday!