20 years ago I worked on perhaps the first entirely self-guided online math curricula to be widely adopted in U.S. elementary schools. While primitive by today’s standards it included a lot of fun activities including one where kids learn measurement by helping some circus fleas attempt death defying leaps over common household objects.

Yes, people thought this would be the end of math education as we knew it…

This might seem quaint in hindsight however, at the time there were people within the publishing company I worked for as well as in the school systems who saw it as some sort of existential threat to math education and an insult to the teaching profession. But in the end the program proved popular and educationally effective, and last I checked there are still math teachers (albeit a shortage of them.)

Fast forward to 10 years ago and I’m attending a conference for the corporate L&D industry. At one point someone got on the stage and started prognosticating about how there would soon be no more corporate training departments or eLearning modules because everyone was going to learn everything from user generated content on YouTube. When I asked the presenter how that was supposed to address organization specific training they basically hand waved the qustion and called on the next person. Today, people do learn a lot from YouTube yet there are still corporate training departments producing eLearning courses.

However, now it’s my turn to play Cassandra telling other professionals in corporate training and K20 education how AI could render the entire notion of training irrelevant. But should my colleagues ignore my warnings the way I largely ignored all those past doomsayers?

L&D has faced “existential threats” for decades, and it’s easy to dismiss the AI revolution as just another false alarm. But this time really is different (no, seriously…) as AI actually can replace much of what traditional L&D does – not just content creation, but delivery, personalization, and even coaching. The question isn’t whether AI will transform L&D, but how L&D will transform itself in response.

Given how most L&D teams are chronically understaffed and overwhelmed, using AI to help with content creation bottlenecks makes perfect sense. Having ChatGPT or a specialized tool like Articulate AI turn your cybersecurity policy into an eLearning course or adapt a generic sales course for each of your company’s product lines are legitimate uses of the technology. In my own work, I recently asked an AI agent to create a customer service role play for an airline based on two similar ones I’d done for hotels and restaurants, and the AI nailed it on the first try!

According to Brandon Hall, 60% of L&D departments are using or actively exploring generative AI, though the overwhelming majority of them (around 95%) were mainly focused on AI for accelerating traditional content creation, with “skill matching” being the next most common use case (35%). However, if using AI to create and/or recommend traditional training materials is all your team is doing, then you might be missing some major opportunities or – worse – automating your own role into obsolescence.

First, our executive stakeholders are not stupid. If leadership sees the L&D department using automated tools to churn out the same old e-learning and PowerPoints – it suggests we could get by with even less budget and / or fewer people.

Second, using new tools to support the old paradigm can blind us to other possibilities. As we’ve discussed at length in other blogs, AI can do much more than generate video scripts and presentation slides. More novel L&D applications include interactive role plays, machine-graded free response assessments, and virtual tutoring tutors / coaches that can answer questions and facilitate open ended discussions within a self guided training course: basically enabling all the learning best practices that many L&D departments previously didn’t have the time or resources to deliver at scale.

Open Full Screen

When the iPhone first appeared, most people thought it was just a phone with a touchscreen – completely missing that it was an entirely new computing paradigm. Similarly, L&D folks who see AI as just “faster content creation” are missing that it’s opening up an entirely new learning paradigm.

I once had a friend – a quirky hippie artist who worked at my local library – who would eat kimchi with every meal. When I asked If kimchi was his favorite vegetable, he said, “No man, I hate vegetables. But fermentation makes the kimchi more nutritious, which means you don’t have to eat as much of it.”

This slightly offbeat diet advice has parallels in workplace learning: as L&D professionals, we’re predisposed to think that training is helpful and people need more of it. But what if people actually dislike training—and AI could help them need less of it?

For decades, knowledge management practitioners have argued that providing people with “just in time” information when they need it is preferable to “just in case” training that someone might quickly forget or never actually use. However, while this is great in theory, most organizations’ knowledge resources are trapped in gnarly intranets that no human can navigate. As one healthcare client explained to us: “We have all these amazing job aids and videos to help doctors and nurses apply best practices, but nobody can find them on SharePoint so they never get used.”

A major focus of our current work is taking these types of knowledge resources and making them accessible to “AI experts”, replacing the SharePoint labyrinth with an expert AI djinni that employees can summon on demand.

Open Full Screen

The idea of integrating learning and knowledge management “in the flow of work” has been a buzzword for years, but with the evolution of expert AI agents, we finally have a chance to make it real. In some cases, AI might be able to provide enough contextual guidance to eliminate the need for pre-training altogether – creating a true “zero training” paradigm.

So, does this mean L&D should give up, fire all the instructional designers, and abdicate workforce training to the IT department? Not at all.

Because, perhaps not surprisingly, AI needs good instructional design just as much as humans do.

Here’s the secret that will save L&D in the age of AI: stop seeing AI agents as tools and start seeing them as your newest, most important learning audience. Your job isn’t just to create content for humans anymore – it’s to educate the AI systems that will work alongside humans.

While there are significant differences in how AI models work versus human brains, there are also commonalities that make instructional design a great foundational skill set for working with AI.

For instance: AI agents experience cognitive overload just like humans do.  While some models can hold millions of words in their active memory, their ability to process and apply that information to solve a specific problem or answer a specific question remains finite. Hence, when you lazily dump an entire knowledge base into an AI without proper structure, organization, and prioritization, you’re essentially creating the AI equivalent of an overwhelmed new employee

Good instructional design principles apply to both humans and AI: Just as you wouldn’t onboard a human by making them memorize manuals, you can’t ‘teach’ an AI agent by simply uploading every document in your organization.  Instead you have to:

• Work with your human subject matter experts to determine what AI agents need to know and do in order to succeed
• Break down complex tasks into logical components
• Determine what information is essential for the AI agent to know at each step of a process, eliminate superfluous content, and provide instructions for how to query supplemental information as needed
• Outline clear objectives and success criteria so that the AI agent is working towards a goal instead of merely answering the user’s questions (especially if the user doesn’t know what questions to ask)

If the above sounds familiar for L&D professionals it should. As our own company has moved from traditional L&D services to creating AI agents, we’ve found that many of the approaches we’ve taken to structuring training content for humans translate surprisingly well to AI agent development.  And we’ve found ourselves acting as intermediaries between human subject matter experts in fields ranging from medicine to forensic accounting and the AI agents we’ve built to assist professionals in those fields, much the same way we work with SMEs on training courses.

While the basic AI models are getting smarter by the day, they only possess the average wisdom of the internet. What organizations need are people who can teach AI agents the proprietary knowledge and processes that give the organization its competitive edge.  And that’s something L&D professionals – with just a bit of AI-related upskilling – are uniquely positioned to provide.

In the early 2010s, marketing agencies who saw social media as just another channel for banner ads got left behind, while the ones who reimagined marketing for Facebook, YouTube, and TikTok became more strategic and influential than ever.

L&D stands at a similar crossroads with AI. And learning professionals need to choose between two possible futures.

Future A: L&D uses AI to create traditional content faster → Content creation becomes commoditized → L&D budgets get slashed → Function becomes marginalized

Future B: L&D reimagines what’s possible with AI → Leads the rollout of AI for human performance support → Becomes strategic partner in AI development → Function becomes MORE valuable

The question isn’t whether AI will transform L&D – it’s whether L&D will seize the opportunity to transform itself.

Hopefully this article offered some useful points to consider for your organization’s L&D strategy. If you’re interested in developing AI agents for workforce training and on the job support, please reach out to Parrotbox for a consultation.