Today’s approach to workplace learning was designed for a different era. The traditional model for learning (even e-learning) reflects the constraints of the age it was built in. At the time, developing learning programs required costly and time-intensive content curation, fixed delivery channels, and no way to know exactly what an individual needed to learn at any given moment for next-level growth.
But now, AI is in the picture.
AI removes most of those barriers, which means the learning model itself is due for a rethink. Learning technology can simply do more now than it could before. There are different constraints to work to, and greater demands for speed and personalization. Learning trends are evolving fast, and it’s becoming increasingly business critical to keep up with them as workforces need to upskill in emerging capabilities like AI.
You can see it in market projections already: The corporate e-learning market is expected to reach $1 trillion by 2030, according to Global Market Insights, and a World Economic Forum report finds that 85% of employers are planning to prioritize upskilling by then too.
But this massive spending doesn’t just represent more money invested in that same, traditional learning model. Learning and development is quickly morphing into something more dynamic and AI-driven than that model can sustain. Something industry experts could barely imagine before. And because skill development might look so different than it does today, I want to paint you a picture of what to expect.
I’ve been building and testing AI learning prototypes in Degreed AI Labs long enough now to have some real conviction about the direction of the market. So here’s my take on where things are headed for 2027 and beyond.
The Formats for Development Content Will Be More Specialized
Today, you can learn by watching a video, reading an article, or attending a seminar. But often, the content doesn’t necessarily correlate to the format in which you’re learning. It’s based on whatever presentation of the content is available. Usually, the formats of most learning content didn’t involve hands-on experience either, even though that’s critical to developing most skills.
Think about what it means to get better at something like presenting. A course can give you frameworks. But a course can’t give you repetitions. It can’t watch you present your actual slides and tell you your energy dropped on slide four or the message you delivered in the data points on slide seven wasn’t clear. You can’t practice the Q&A with a skeptical audience over and over until you get it right.
When the format of the content matches the task, and the experience is built around what you’re actually trying to do, something changes. The feedback is more specific. The practice gets more useful. Suddenly, experiential learning can be done on demand and in the moment of need for the exact situation you’re trying to prepare for.
Personalization Will Mean Dynamic, Responsive Learning
We’ve been designing courses for an imaginary average person for decades. First, it was one-size-fits-all—the same course for everyone. But even the tailored or adaptive learning that’s emerged since then takes the form of pre-built paths that were created in advance and based on certain assumptions about the direction or speed of someone’s learning. Better than the first model, but not truly “personalized.”
The future looks more like a conversation. One that starts by understanding what you already know, adapts the content when you’re confused, goes deeper when you push back, and checks in on what you need next, rather than continuing on a predetermined track.
I tested this kind of dynamic course experience in AI Labs to see exactly what it might look like in practice. I created an experiment for a new kind of development course that is designed for easy daily learning. The content wasn’t created before I started the conversation. Instead, the AI persona was just ready for me, primed with the context of my upskilling history and proficiency.
In the experiment, the first screen was a selection of topics based on my previous learnings and interests. I selected a fairly technical topic in the demo, one where I had some existing context and skill history in place. The difference was immediate. It started with mid-level content that matched my proficiency level. When I asked a question that took things in a different direction, it followed without friction.
That’s what personalization actually means when it works: A real-time response to a real person who has dynamic needs and shifting focus.
Expect Specific, Situational Learning Content
The third trend might be the most significant and the least discussed. The moments when people most need to learn aren’t scheduled. They happen two hours before a hard conversation. In the middle of a project that’s going sideways. Right before a presentation to someone who knows more about the topic than you do.
Traditional learning can’t support those moments at the level of specificity people need at work. By design, it exists outside of them. What AI makes possible is learning that’s situational, meaning that it’s generated on demand, around the specific scenario you’re actually facing, and quickly enough to be truly useful.
Imagine describing a high-stakes call you have coming up and getting a realistic practice scenario back within minutes: multiple personas with distinct personalities and questions, the ability to share your screen if you need to demo something, live feedback, and a debrief at the end. Not a generic communication skills module. A specific rehearsal for the exact actions you’re about to perform.
I created a prototype specifically to meet this kind of need. It would allow you to specify, for example, that you were going into a meeting with three role-specific stakeholders to try to get buy-in for a defined initiative. Within the model, you’d then enter into a practice session where you would present, and AI would play the stakeholder parts, then ask questions and provide feedback related to their perspectives on the presentation by function and role.
Suddenly, preparing for that call that’s in an hour is more doable and effective.
The Future of AI-Powered Learning in 2027
What we think of as AI-powered learning today is nothing like what it will look like in 2027 and beyond.
The individual tools are becoming genuinely impressive. But even right now, the best learning experiences require people to seek them out. You have to know the resources exist, know they apply to your situation, navigate to them, and set them up. And they still are not totally personalized and situational to the moment of need.
The version of this that I think will actually work is one where the need and the experience find each other. Where an employee can describe what they’re facing conversationally, or where a manager can describe what a team needs, and the right experience can be automatically created in real time, contextually, without any added infrastructure work required to produce the solution.
Learning is moving toward experiences that are built for the task, not for the average person. Built for the individual, not for the masses. And available at the moment of need, not scheduled six weeks in advance. That’s the direction innovation is going. And it’s closer than most people think.
I know because we’re already prototyping the technologies that can do it.
Still, I want to be careful here, because this space moves fast and the hype is real. The experiences I’ve described above are prototypes today. Some will make it into products. Some won’t. But the technology still has rough edges, and making any of this enterprise-ready takes more than a good demo.
If any of this connects to what you’re thinking about and you’d like to talk about it, I’d welcome the conversation.
Everything discussed in AI Labs content is experimental, in progress, and not guaranteed to become part of the Degreed product suite.
Unser heißer Draht zu Ihnen.
Exklusive Einblicke, Infos zu bevorstehenden Veranstaltungen, Updates zu Degreed-Lösungen – das alles erwartet Sie in unserem monatlichen Newsletter.