By Bianca Raby
Do you know what is the most widely used learning platform on the planet right now? It is not your favourite LMS. It is not a corporate intranet. It is not the carefully sequenced e-learning module your team spent three months building.
It is YouTube and guess what? Nobody in L&D built it.
That fact is either mildly embarrassing or deeply instructive, depending on how you look at it. I’d argue it’s the latter. Because YouTube didn’t win by being the best-designed learning environment. It won because it met people at the exact moment they needed to know something and got out of the way. That is learning in the flow of work in its purest form (Bersin 2018). Unscheduled, learner-driven, and ruthlessly just-in-time (as long as you have an internet connection).
So I ask ourselves as L&D professionals, why are we still designing against that instinct?
Most formal training is built around a simple idea that if we schedule people to learn something, they will learn it. We block out time, build a course, run a workshop, and assume the transfer will follow. However, the research on this has been consistent for decades that it largely doesn’t.
The transfer doesn't now happen generally because people don’t care, but because there is a significant difference between learning that happens in a specific context and learning that happens in anticipation of context. When someone pauses mid-task to find out how to do the next step, the stakes are real, the motivation is high, and the application is immediate. When someone attends a workshop on the same topic three weeks before they need it, they are largely pre-loading information into a container that will be mostly empty before it is ever used.
Herman Ebbinghaus mapped this out in the 1880s. The forgetting curve has not changed just because we updated our authoring tools (Ebbinghaus, as cited in ScienceDirect n.d.).
Learning in the flow of work is not a new concept. It is actually a very old one. Apprenticeship, mentoring, on-the-job coaching all existed long before the training department did. What is new is our ability to design for it intentionally and at scale and I will argue that in 2026, with the tools available to us, there is little excuse for not doing so.
Over the last decade, L&D has enthusiastically embraced microlearning and rightly so, to a point. Short is considered generally better than long and focusing is better than sprawling. However, somewhere along the way, microlearning became shorthand for ‘cut the old course into smaller pieces and call it new’ (Shank, 2018).
Short content is not the same as contextual content. A three-minute video on a topic someone doesn’t currently need is still three wasted minutes. The consideration should not just be about how long the video is, but about when it appears, and is the learner ready for it right now.
True flow-of-work learning has a few characteristics that most microlearning catalogues do not. Such as;
It surfaces at the point of need
It is specific enough to be immediately actionable
It requires something of the learner, for example: a response, a reflection, an application before they move on, and
It connects to what they already know, rather than assuming they are starting from scratch.
These are design choices, not just content choices and they require L&D to think less like publishers and more like architects of experience.
Consider what happens when organisations try to roll out AI capability building right now. The instinct is often to send people to a course. Bring in an expert. Run a half-day workshop to ‘tick the box”.
The problem is that AI literacy is not a static body of knowledge. The tools are changing weekly. What someone needs to understand today may not be what they need in four months and the concepts that matter, such as what bias looks like in an AI system, when to trust a result and when to question it, these are best understood in relation to actual use, not in the abstract.
This makes AI literacy an almost perfect test case for flow-of-work design. The learning cannot be front-loaded, because the landscape will have shifted by the time someone needs to apply it. It cannot be purely conceptual, because the ethical and practical judgements involved when using AI only become real when in context. And it cannot be one-size-fits-all, because the risk profile, use cases and starting points vary enormously across roles and industries.
What works instead is learning that is embedded in the actual practice of using these tools. Content that combines foundational understanding with structured reflection and is designed to be stacked together over time rather than consumed all at once. The goal is not to make people experts before they start. It is to build judgement alongside experience.
The same logic applies well beyond AI. Any rapidly evolving capability area like data literacy, cross-functional collaboration, change leadership is better developed through repeated, contextualised encounters than through a single event.
Designing for the flow of work does not require an entire overhaul of your L&D approach. It does require a genuine shift in how you think about what learning is for and where it belongs. Here are some principles worth building around.
Start with the moment of need, not the topic
Before designing anything start by asking when in a person’s actual workflow does this knowledge gap appear? What are they trying to do when they hit the wall? The answer shapes everything from the format, the length, the tone, the entry point. If you cannot answer that question clearly, you are probably designing for the wrong moment.
Build in retrieval, not just delivery
The research on retrieval practice is robust. Being asked to recall or apply something is significantly more effective for retention than being exposed to it again (Brown, Roediger & McDaniel 2014; Roediger & Karpicke 2006). Even a short prompt at the end of a three-minute module. Something that asks the learner to make a decision, write a sentence, or identify an example from their own context will dramatically increase the likelihood that the learning will stick. Most microlearning skips this entirely. Do not skip this.
Design for stacking, not for completion
Instead of building a course that someone completes once, design units that can be returned to and built upon. A five-minute module today, a slightly more complex application next week, a reflection exercise after a real use case. The goal is not a single learning event but a cumulative building of capability over time. This is how expertise actually develops, and it is the model that flow-of-work design is uniquely positioned to support.
Let the learner drive where you can
YouTube works partly because nobody assigned it. People go there when they want to know something and that autonomy matters (Bersin, 2018). Wherever possible, design learning resources that people can find and use on their own terms, not just resources they are pushed through by a system. This might mean building a well-tagged resource library instead of a mandated pathway. It might mean making a series of short modules available for self-selection rather than rolling them out in a set sequence. Learner agency is not a nice-to-have. It is part of what makes learning stick.
Keep it under six minutes per unit, but make every minute count
Short modules are only valuable if they are genuinely focused. So, just one idea, clearly explained, with an immediate application or reflection prompt. If you find yourself trying to squeeze three concepts into four minutes, you have two more modules to build.
None of this is technically difficult because the challenge is more structural and cultural than it is a design problem.
Often, it requires a more honest conversation about the role of the course itself, because a course is one solution to a capability problem. It is not always the right one, and it certainly should not be the default it has become in most organisations. When someone is underperforming, the instinct is often to send them to training. When a team is misaligned, the instinct is to build a workshop. But capability gaps have many causes like unclear expectations, poor feedback loops, broken processes, low psychological safety.
So a well-designed learning experience alone cannot fix a management problem or a systems problem. L&D professionals have known this for years and the performance consulting literature has been making this argument since the 1990s. What has been slower to change is the organisational expectation that training is the answer, and the willingness of L&D teams to push back on that when it isn't.
Designing for the flow of work is partly a design shift and partly a positioning shift. It requires L&D to show up differently in the organisation as less of a production team that fulfils course requests, and more as a genuine partner in diagnosing what is actually getting in the way of performance.
Flow-of-work learning also requires L&D to give up a certain amount of control over when and how learning happens. It requires a willingness to measure differently by focusing less on completion rates and giving more attention to whether behaviour is actually changing. It requires stakeholder conversations that are harder to have, because the value is longer-tailed and harder to point to on a dashboard.
Because remember, the people in our organisations are not lacking exposure to information. If anything, they are drowning in it. What they need is learning that is sharp enough to be useful precisely when they need it, and designed with enough respect for their time and intelligence to get out of the way as soon as it has done its job.
YouTube is not going anywhere. Nor is the underlying human instinct to seek knowledge at the moment we need it, in the most direct form available.
The opportunity for L&D is not to compete with YouTube. It is to understand why it works so well and to bring those principles into the design of learning that is actually embedded in the work that matters. Short, contextual, retrievable, learner-driven, and stackable over time.
We have always known that people learn best when the stakes are real and the application is immediate. We just haven’t always been willing to design around that truth as it's harder. But maybe it’s time we did.
Bersin, J. (2018). A new paradigm for corporate training: Learning in the flow of work. Josh Bersin. https://joshbersin.com/2018/06/a-new-paradigm-for-corporate-training-learning-in-the-flow-of-work/
Shank, P. (2018, updated 2025). Microlearning, macrolearning: What does research tell us? eLearning Industry. https://elearningindustry.com/microlearning-macrolearning-research-tell-us
ScienceDirect. (n.d.). Forgetting curve. Elsevier. https://www.sciencedirect.com/topics/agricultural-and-biological-sciences/forgetting-curve
Brown, P. C., Roediger, H. L., & McDaniel, M. A. (2014). Make it stick: The science of successful learning. Harvard University Press.
AITD offers several courses on both AI and eLearning:
AI Essentials for L&D Professionals: Do you want to transform your L&D workflow and enhance learner experiences using Generative AI? In this blended learning course, you’ll learn how to use Generative (Gen) AI for a range of key L&D functions, including skills gap analysis, content creation, personalised learning and feedback. Register now.
eLearning: Foundations: This is Part I of an engaging, social learning suite of courses that provides you with access to learning experiences, activities and a comprehensive knowledge base. Register now. You may also be interested in eLearning: Planning and Design; and eLearning: Production and Delivery.
Bianca Raby is the Founder and CEO of Oppida, a learning design and capability building agency based in Adelaide. She works with educators, organisations and leaders across education and business to rethink how learning is designed and delivered, with a strong focus on building capability in AI literacy and preparing teams for a rapidly changing work environment. Her work sits at the intersection of learning, technology and the future of work, where she explores how individuals and organisations can better navigate complexity and change. Bianca is the author of the LinkedIn Newsletter and Substack publication Learning, Tech & the Future, shares insights on her YouTube channel, and is the author of Online Learning: So, you want to be a learning designer?