In October, we held our virtual user summit, Rustici: The Gathering, which also happens to be my favorite event of the year. For the second year in a row, we hosted a panel of AI and eLearning experts. I was really excited to hear about where L&D is in regards to AI, what’s been successful and what’s coming next.
Here are a few nuggets and highlights from the session that the audience and I really found interesting, but be sure to check out the full recording for all the insights and full answers.
AI in the present – Where are we seeing AI have the biggest impact in learning and development?
Megan Torrance, TorranceLearning
There’s a lot of activity in personal productivity for L&D professionals, including a lot of generative AI chatbots and some text-based and media use. Data analysis is also being impacted, though it still requires a fair amount of human support and supervision.
Vendors are also discussing role-play chatbots or coaches that help sales, leadership and customer service professionals to practice conversations that can be difficult to do with other people. This practice can be done anytime, anywhere and repeatedly while providing good feedback.
And in eLearning standards
Stephen Kalnoske, Rustici Software
In the compliance space, vendors are now baking AI into their Learning Management Systems (LMSs), tools or learning experience platforms. The long-term dream of getting the right training out to learners at the right times is now starting to be delivered by AI in these platforms. Practically, this drives down the corporate cost of training. Previously, it was easier to just give everyone all the compliance training, but AI tools can now slice the training in an automated and smart way to give learners what they need. And vendors are incorporating this into their value proposition.
Can you share a project or where AI was used to create a more customized or efficient training experience?
Reed McLean, Rustici Software
With Rustici Generator, it looks at the content, parses out the content and generates metadata about it. This makes it easier to identify and describe a huge content library. It also makes it possible to find a specific regulation or phrase for when rules are updated or the content needs updating based on changes.
Stephen
I worked on a product called Tutorbot, an AI tutor given to students in an intense seven-day cybersecurity training course, and who weren’t in their offices. The purpose was to prevent students from falling behind, as there is no catching up in that kind of environment. Tutorbot allowed students to take a test on what they learned, which identified their weak points. The training was then redelivered in “bite-sized parts” focusing on those weak points to help them catch up and succeed in the course.
Before AI, a lot of training created in an authoring tool had to be a multiple-choice quiz because of the need to grade it. With Tutorbot, they used multiple choice, but they also asked students to explain concepts in an empty text box. The grading was no longer binary, and the system could offer feedback like “you’re close” or “you just missed this little nuance thing,” which opened up a whole new area of grading and feedback.
Megan
I love that so much! I love that they’re able to interact with Tutorbot “off hours” since the facilitators need to go home and sleep. A startup in Ann Arbor had an AI teaching assistant that could answer questions. They found that the primary activity times for their tutor were between 11:00 p.m. and 3:00 a.m., hours when faculty were not available. These are the “golden cramming hours” from college and make these applications powerful. They also support moments of need where someone can pull up a coaching bot instead of calling their instructor.
At DemoFest last year, we gave learners an empty text box instead of a multiple-choice question, which is the standard way to score data in SCORM and xAPI. Instead of an interactive roleplay opposite the learner, the tool gave feedback on what the learner did, and then they could try again or move on. It’s really powerful for generating retrieval practice, where the learner has to pull the correct action from memory. Instead of the traditional binary “you got it right,” the learner receives feedback on what was right and what was not, and they are given the opportunity to try again.
AI in the future – Where is AI heading in the eLearning and training space over the next three to five years?
Reed
A really interesting, yet challenging, use of Large Language Models (LLMs) for Rustici Generator is using them in non-interactive contexts. Since a lot of LLMs are built for a chat back-and-forth interface, using them to design functions with fuzzy rules and fuzzy inputs while still trying to get discrete values is difficult. We’re using LLMs to parse content from any eLearning course and receive concrete information from it. Generator uses asynchronous jobs where you “start it up, let it go and it comes back with some words.” This will be interesting when working with problems that don’t have a strict set of parameters. These problems are not really solved yet.
Megan
There is an opportunity to personalize, which is a key theme. You get personalization in social media, human lives and when talking to managers, so why should you take the same course as thousands of other people? Personalization is the most interesting and effective opportunity because it taps into efficiency and adult learning theory. Connecting training to what the learner actually needs and when they need to apply it is important. The challenge for organizations will be ensuring everyone is getting the consistent story and the same framework so they have reasonable expectations of a person’s skill sets.
Stephen
For compliance professionals needing to ensure accuracy and defensibility, the data part of AI is not quite there yet. Personalization comes down to the ability to collect data about interactions and people, and the challenge is what to do with that data and how to get it into customized and personalized courses, which xAPI is a great answer for. The defensibility of what is put into and produced by the AI is easier in the compliance space because the content is not “bespoke secret sauce.”
Whew! Megan, Stephen, Reed and Brian covered a lot of AI information during their full Rustici: The Gathering session. If you have questions about how AI fits into your learning ecosystem or how Rustici Generator could help fuel your own AI tools, reach out and ask us anything!
