I wrote the article “The Path to AI” for the January 2020 edition of TD Magazine. The purpose of the article was to provide a foundation for how L&D pros can begin to approach the application of AI within workplace learning.
Rather than get enamoured with specific technologies or marketing hype, L&D must seize this opportunity to rethink a lot of how we do what we do. No, we can’t stop the already-moving train that is workplace learning. But, to get the most value from this new technology, we have to do a bit of a mental reset and ask ourselves “how would we do this if we were just getting started today?” L&D largely failed to do this during the emergence of other inflection point tech, such as social and mobile tools. We know what happened there.
The AI story is decades old, but the idea of applying machine learning to help people do their jobs better is in its infancy. Therefore, there isn’t a lot of reference information that applies specifically to L&D. I curate an online magazine on AI in the Workplace, but the examples typically focus on operational or general human resource applications. Within L&D, AI is most heavily mentioned in relation to marketing specific products (that may or may not actually use AI). Overall, we are already playing catch up when it comes to understanding AI, but this certainly isn’t new for our field.
Back to the TD article …
Because my intent was to provide some perspective for how to fundamentally think about AI as related to learning and performance, the article was purposefully light on specifics. Frankly, there just aren’t many readily-available examples or stories of meaningful AI application. I also tried to avoid direct references to my work at Axonify, as I don’t want my industry contributions to be seen as advertising. I would rather L&D pros start working in earnest with their internal partners to discover potential applications for AI in their work than get distracted by vendors.
Then I received the following common regarding the article …
I wish this article had links to specific use cases of AI in use by TD. This article makes the point that it’s our future… but then repeats it redundantly without casting a vision nor giving us a place to start exploring the possibilities.
So I figured I would take the opportunity to add some more detail to the conversation via a follow-up blog post. Here’s a list of 12 use cases for AI in workplace. All of these applications are real, but they are also in varying states of adoption given the iterative nature of workplace technology implementation.
Tagging should be familiar to any L&D pro that administers a content management system. In order for the system to properly categorize and deploy content, it must understand what the content is about. Therefore, we apply meta tags to each piece of content. Historically, this has been done manually. For example, I tagged this article with “ai” and “technology” to help my website sort it appropriately. Today, content systems are increasingly using AI to automatically determine and apply these tags. This is true of structured content, including job aids and video transcripts, as well as user-generated content, such as social posts. The AI application scrapes the content to find keywords and automatically associates them to the content to support search and enrollment. This is a basic AI use case, but it has the potential to improve outcomes related to tagging and reduce administration as content is uploaded and maintained over time.
How much time does your L&D team waste on administrative tasks in your LMS? This is where AI can come into play as a foundational capability within learning technology. This is not a heavily implemented use case quite yet, but AI is already able to handle repetitive administrative tasks that do not require human decision making. This includes tasks such as topic assignment, course scheduling and enrollment, report generation and content management. Every learning technology provider with an R&D function is exploring how to leverage AI-enabled automation to improve the administration experience within their platform.
Right now, this is what too many people think about when AI is mentioned in L&D. While some chatbot applications are already being used to support programmatic training and coaching (ex: Mobile Coach), this tech has been more heavily adopted in general HR practices to support employee self service. Given the conversational limitations of modern AI, most chatbots are a combination of a pre-programmed decision tree and natural language-based search. The bot may apply machine learning to improve its responses and recommendations over time, but it’s not actually talking to the user (yet).
Translation is a major challenge for L&D teams, regardless of business size. If your organization operates in just Florida, you likely have to provide employee information in at least 3 languages. Translation is time-consuming and expensive. The inability to translate enough content quickly puts large groups of employees at a disadvantage. AI should be able to help us solve this problem very, very soon. Have you used Google Translate? This is a good indicator of where the bar currently rests with AI-enabled translation. It’s very good, but not quite there yet. Some content development teams are using AI-enabled translation to create a first draft, which is then reviewed by a native speaker to ensure details and cultural specifics are accurate. Within a few years, learning technologies will likely include AI-enabled translation as part of their content development toolset.
In the early days of the internet, search was extremely specific. If you didn’t enter the exact keyword, the system would not be able to find your content. Today, consumer search engines use natural language processing to associate your keywords with related content. AI is used to break down your search request and provide the best possible input to the associated information database. This is why you can speak to Alexa in a conversational manner without using specific command terms. AI converts your speech to text and then executes the search based on its contextual understanding of your request. Again, it’s not perfect, but it’s evolving quickly. And, because many workplace applications leverage cloud-based search technology, this capability is making its way into learning platforms.
L&D spends A LOT of time and money on content. Whether it’s built internally or sourced from a provider, content represents a major chunk of an organization’s L&D investment. At the same time, we never seem to be able to build content fast enough. And we already know that generic, off-the-shelf content isn’t enough to help people solve nuanced performance challenges. Enter AI. Machines are already able to generate structured content from source material. It’s not the best content, but it’s pretty good. AI is also getting scary-good at generating media content (ex: Generated Photos). L&D teams are starting to apply AI to generate simple content, such as articles and questions, that is then reviewed and improved by a human developer (ex: Wildfire). In the near future, L&D will finally be able to keep pace with business needs and provide highly-contextualized content – when it’s the right solution to the problem.
This is another use case you already hear a lot about in L&D. If you have a massive course catalog with thousands of items, how do you help employees discover high-value content without having to know what to search for? Well, you build the Netflix of Learning, of course! The LXP category is based on their ability to apply data and machine learning (sometimes) in order to aggregate and recommend content to users based on their past interactions. In other words, if you experienced/liked this course (or people with similar profiles to you experienced/liked this course), you may also like these courses. However, this approach to recommendation is fundamentally limited. After all, how often does Netflix suggest something you really want to watch? The system just doesn’t know enough about you. L&D must improve its data practices so AI has more to work with. By expanding a user’s data profile to include not just content consumption but also changes in knowledge, behavior, confidence and performance results, L&D will be able to use AI to recommend a variety of proven solutions – not just courses.
Most coaching is a combination of firefighting and dart-throwing. Managers react when a clear issue arises and make informed guesses regarding the causes and solutions to performance problems. It’s just difficult to track the nuanced, day-to-day performance of every employee on the team, especially if you don’t actually work alongside them on a regular basis. Improved data practices along with AI-enabled technology can remove a lot of the guesswork in coaching. Manager reporting is started to transition from flat spreadsheets to three-dimensional insights with actionable recommendations. AI can help managers recognize potential issues early, before them become problems that impact business performance, and recommend proven actions that are more likely to result in positive outcomes.
L&D’s role needs to become more about systems and connections, less about content and creation. AI can enable this transition, especially when it comes to connecting people who “know” with the people who “need.” In a large, complex organization, employees often rely on people within their direct circle of influence for support. Unbeknownst to them, better resources likely exist within other teams or other regions around the world. AI-enabled technology can connect people with performance challenges to peers who overcome similar challenges and may therefore be a good match for mentorship.
We’ve already addressed content authoring, translation and recommendation, all of which are parts of the overall personalization story. Just as it does in consumer technology today, AI can further personalize the workplace experience in a variety of ways. For example, with Axonify, employees experience personalized training every time they log in, as the application balances personal development needs with organizational priorities and always delivers the right-fit learning experience.
You know how L&D can never determine how their solutions actually impact business results? Yeah, AI can help fix that. However, like many of the use cases on this list, it all starts with data. L&D must evolve their strategy to include data-rich tactics that paint a more holistic picture of an individual’s performance. Machine learning is then applied to determine how specific learning solutions do (or do not) trigger incremental changes in knowledge, behavior and performance. This insight can be used to develop a business case for added L&D investment, make smarter decisions regarding learning initiatives and further enhance personalization efforts.
Skills are a big topic of conversation in L&D right now. Organizations are rapidly redesigning jobs as they apply new technologies and work to remain competitive. L&D must get ready to support large-scale reskilling initiatives. But on which skills should an organization focus? Who already has these skills? And what are the best ways to help others with potential develop these skills? Data and AI can help address this challenge at the speed and scale required. Rather than chase the changing needs of the workplace, AI-enabled technology can identify potential gaps in organizational capability. It can then automatically adapt the learning experience for employees who will need to develop specific skills as part of future job opportunities. The only way continuous reskilling can be accomplished at the speed and scale a complex business requires is through a lot of data and the right AI.
By no means is this a comprehensive list. I’m sure there are plenty of ways data and AI can be applied to enable employee performance that I have not yet considered. But everything on this list is real and actively being worked on today within HR and L&D. This is where we’re going. It’s time for L&D pros to start talking about what it will mean to work in an AI-enabled workplace.