Unless you’ve been hiding under a rock for the past few months, chances are you’ve caught one or two (or seventeen) conversations about artificial intelligence (AI) and ChatGPT.
It’s a hot topic thanks to recent public releases of some deeply impressive (and for some, concerning) AI systems. In the art world, prompt-driven AI tools like Midjourney can create stunning visuals on demand. And text-based ChatGPT can output words, poems, essays, and even decently written code.
AI and machine learning (ML) strategies have been quietly employed in many pieces of technology for decades. But this explosion of public interest marks a potential turning point for society as more and more people begin to grapple with the growing reality—and approaching ubiquity—of AI.
Learning and Development (L&D) is no exception. Docebo’s CEO and Founder, Claudio Erba, tackled the subject in a recent webinar, Generative AI and ChatGPT: Reshaping Online Learning. Along with Fabio Pirovano (CPO and Co-Founder, Docebo), and Massimo Chiriatti, (CTO/CIO, Lenovo) Claudio explored the macro trends in L&D and exactly how emerging AI technology will play a key role, as well as some of AI’s inner workings, its risks, limits, and more.
You can watch the full recording here, but for a six-minute recap, read on!
How AI will reshape online learning
There are two key macro trends that L&D professionals need to keep an eye on (whether they’re interested in AI or not):
- Employable people are decreasing as a percentage of the population
- Overall productivity in the workforce is declining
Consequently, employee retention and talent acquisition strategies are more important than ever. Keeping people around longer, increasing their productivity, and attracting top talent have always been key, but when you take these macro trends into account, the stakes are even higher.
L&D teams are responsible for improving employee retention and productivity, as well as talent acquisition. That’s not new. What is new is how AI can help them achieve these goals. L&D teams can leverage AI and machine learning tools to improve operations in the following ways:
- Automation of tedious tasks, freeing up administrative time and improving accuracy
- Hyper-personalization of L&D material, matching people and needs faster and seamlessly serving multiple distinct audiences
- Content generation that helps reduce overhead and improve accessibility
- Conversational business intelligence, providing quicker and better access to insights
Examples of AI’s increasing role in improving L&D operations
Automation
Imagine doing the same thing over and over and over. Without a break. All. Day. Long. While it might seem boring to us, it’s what computers were designed to do. The first major commercial uses for AI and ML were generally in automation, and will likely lay the foundation for adoption for years to come.
In L&D specifically, automation via AI might look like:
- Auto-tagging. Classifying 12 pieces of content is no problem for a human. But what about 12,000? What if the classification parameters need updating on a regular basis? This is a job for an algorithm.
- Transcribing. Video-based learning is huge, and with the TikTok generation on the rise, we can only expect the amount of video to increase. But manual transcription takes time. Modern AI can perform video-to-text transcriptions with impressive accuracy and speed.
- Translating. Google Translate works in a pinch, but if you need to convert a full course into multiple languages, you’ll need more firepower. AI can create very good first drafts for native speakers to review, saving a lot of time and money.
- Search indexing. Making things easy to find is no small feat when dealing with data and content. Social learning can be especially messy, since a lot of the value is created organically with minimal structure. AI and ML tools can work in the background to create order out of chaos, reducing the time needed to find the information a human (or another system) needs.
Automation is more than simply time-saving. It is an optimizer of human attention and skill. Most people don’t enjoy mundane, repetitive tasks (and we can get tired and make mistakes). Offloading the digital grunt work to machines frees people up to tackle more creative and strategic challenges.
Hyper-personalization
One-size-fits-all solutions aren’t created because it’s what people want. They’re created because they scale. In L&D, this looks like one instructor trying to manage 35 students. Or 3,500 employees completing the exact same security training.
Most people would benefit from 1:1 coaching and mentoring vs the “crowded classroom” approach. But that’s expensive. Often prohibitively so.
AI and ML can help reduce the scaling issue via hyper-personalization. It can’t get to the level of 1:1 coaching (yet), but it can help filter out a lot of the superfluous and irrelevant content a learner doesn’t want or need, and it can adapt to user behavior and dynamically adjust in response.
Here’s an example of how this could play out in L&D:
A UX/UI developer is navigating her company’s L&D system. Based on the completion rates of previous courses, her title, the activities of her peers, and her own self-defined goals, the AI recommends a custom-built curriculum. As she moves through these courses, the AI detects a bias for a certain kind of programming language (e.g. time on articles, assessment results, etc.)
Depending on the programming of the AI, the bias could either be reinforced (show her more courses that use this language) or be countered—depending on the overall L&D strategy. Or maybe nothing is done about the bias and it’s simply reported as an observation.
Content generation
This topic has been getting a lot of attention lately.
Tell a good art AI to draw a picture of a cat in the style of Picasso (an example used by Massimo Chiriatti during the webinar), and you’ll get what you ask for.
You can ask ChapGPT to give you 7 blog topics on the biggest challenges in professional development (another prompt used in the webinar) and it’ll get good stuff to you in a few seconds.
L&D content developers would do well to understand how they might leverage these tools to enhance (not replace) their output. During the webinar, Docebo’s CPO and Co-Founder, Fabio Pirovano, gave a quick demo of Docebo Shape—a solution that allows users to fast-track L&D content using AI tools.
With just a simple prompt, you can create an explosion of content. It goes something like this:
You prompt ChatGPT to write a short article on “the most important things a new UX/UI designer needs to know.” AI generates an article in seconds. You review the AI output to fact check, tweak, and polish language, then you feed the content into Docebo Shape and define output length. Docebo Shape AI creates an engaging deck in a few minutes. You review the deck (and tweak the design and content if needed). Then you use Shape AI to create audio voiceover for the deck (to increase accessibility) and to translate the deck into multiple languages.
From a single prompt, you now have several decks in multiple languages with audio voiceover files to go with them. If you did this job manually, it would take days (or longer, in many cases). But with AI, it takes minutes. And the best part is, aside from the initial prompt and a few quick reviews, AI does the bulk of the work.
Conversational business intelligence
Reporting, dashboards, analysis… These are the staples of business intelligence (BI) and underpin many major decisions.
Conversational BI refers to using AI prompts to crunch numbers and surface visual data easily, cutting down on time wasted asking for and waiting on reports.
An L&D administrator could use conversational BI in the following way:
They could ask an AI tool to show them course completions for last month in all major categories except for security training. It would spit out an accurate table. The admin could reformat the table into a bar graph or pie chart with a single click, or even ask the AI tool to extrapolate the data and compare it to last year’s results.
(Of course, for all of this to work, the underlying database needs to be well maintained and accurate, which is yet another job AI can help with!)
The risks of AI for L&D
This all sounds very futuristic and convenient. (Too convenient even.) And it has people asking, “What’s the catch?”
This question is explored at length in the webinar, but here’s the gist:
- AI is not conscious and has no sense of ethics or morality. It will simply do what its programming dictates, and that programming is written by fallible humans. Blindly trusting AI outputs is a mistake: There should always be a “human-in-the-loop” to scrutinize.
- AI can amplify biases, prejudices, and other inequalities. We already see this happening on social media platforms. If an algorithm is designed to learn what content “engages” you, and it doesn’t differentiate between “happy engage” and “angry engage” then it may well “optimize” towards filling your feed with content that angers you (with no regard for whether or not the content is true or good for you).
- It’s a disruptive and relatively new technology. This always carries risks—we’re not sure exactly how it will evolve, how it might be regulated, etc. Unintended consequences are the norm with fresh tech, so a healthy dose of risk management should accompany any AI/ML strategies.
Remember: Artificial Intelligence is a discipline that uses computer systems to read the footsteps of our past in order to suggest and generate future steps.
Want to learn more?
If you want to dive deeper into these topics and get more insights from Claudio, Fabio, and Massimo, you can watch the full recording of the webinar, Generative AI and ChatGPT: Reshaping online learning on demand here.