In my 20 years of being an L&D consultant, I have spoken with hundreds of L&D leaders, administrators, learners and business stakeholders, across every sector and every size of organization. And I have almost never met anyone who is genuinely happy with their LMS. Not the people who chose it. Not the people who run it. Not the people who are supposed to learn from it. Not the people who pay for it.

That is a remarkable realization when you stop to think about it. An entire category of software that almost nobody loves, that we keep buying, replacing, and buying again. When a whole profession is quietly dissatisfied with the same tool for a few years and then wants to replace it. The problem is not the tool. The problem is what we are asking the tool to be.

This newsletter is the first in a new series about where L&D is heading in the next years, and it is a radical, detailed and pragmatic expansion of the vision series that came earlier this year built around a single personal belief: I believe the future of L&D is agentic. Not a better platform. Not a smarter LMS. A layer of intelligence underneath everything, on which AI agents run the actual work of learning. From spotting a capability need to designing a response, to checking whether it worked. Traditional learning software, the kind we have spent decades configuring and complaining about, largely goes away.

That is where we are going. But to see why it is coming, we have to start with an honest look at why the thing we have today makes nobody happy.

The LMS was not a mistake. It did exactly what it was designed to do.

It was built in and for a different era of corporate learning, when learning mostly meant courses, and courses mostly meant classrooms and, later, e-learning modules. In that world, you needed something to administer it all. To enroll people. To schedule sessions. To record completions. To prove, for compliance and audit, that the required training had happened.

The LMS is very good at that. It is, at heart, an administration system for formal learning. A system of record. And for the era of learning it was built for, a system of record was enough.

The problem is not that the LMS does its job badly. The problem is that the job changed, and the LMS did not.

Look closely at who is unhappy, and you see the same underlying issue from four different angles.

Learners are unhappy because the LMS asks them to behave in a way that no longer matches how they work. The interface is clunky. The catalogue is overwhelming. Search returns everything except the thing they needed. And most of all, they are tired of being told to "go to the LMS" or "go to the LXP" whenever they have a question. When you need an answer in the flow of work, being sent to a separate platform to find and complete a course is the opposite of help.

That learners are not happy with their LMS/LXP is nicely illustrated by Jane Hart’s annual Top 100 Tools for Learning. This year is its final year (will anyone take over?) and to nobody’s surprise, the top 10 does not contain a single LMS or LXP. Even more, the first ‘learning tool’ in the list is Kahoot, which is more a quiz or engagement tool. Then there is Articulate (a content authoring tool) at number 25. Moodle is at 69 the first tool that can be referenced as being somewhat of an LMS (more a course management system). And I find that a rather painful observation

Administrators are unhappy because they carry the hidden weight of keeping the system running. Mostly manual maintenance, uploads and updates. Courses and programs that have to be created manually (many based on excel input sheets!), and learning paths that need to be curated manually. Integrations that break. User records that do not match across systems. Reports that take days to produce and still do not answer the question that was actually asked.

L&D leaders are unhappy because the LMS cannot help them with the thing they are increasingly asked to do, which is to prove that learning makes a difference. The LMS knows who completed what. And compliance reporting tells you who is overdue. But the LMS does not know what changed as a result, if anything changed at all. It was never designed to connect learning to performance, so it cannot.

And the business is unhappy because it pays for all of this and rarely sees the value. It sees a cost line and a completion report, and it quietly concludes that learning is an expense to be managed rather than a capability that drives results.

Four groups. Four different complaints. One root cause. We are using a system built to administer learning as if it were a system meant to enable learning and to understand it. It cannot be those things, because it was never built to be.

This is why the endless search for a better platform never ends well.

When we are unhappy with the LMS, we assume the answer is a different LMS, or an LXP bolted alongside it, or a new content library, or a fresh integration project. We run a selection process. We migrate. We reconfigure. And within a year or two, the same four groups are unhappy again, for the same reasons. And meanwhile the number of tools available in the market has grown so big that making a choice is only getting more complicated.

We keep solving the problem at the level of the tool. But the dissatisfaction does not live at the level of the tool. It lives one level down, in what we are asking the whole model to do.

The LMS was designed to be the center of gravity for learning. Everything was supposed to live in it, run through it, be measured by it. But learning stopped living in one place a long time ago. It happens in the flow of work, in conversations, in tools scattered across the organization, and now, increasingly, in exchanges with AI. Asking a single platform to be the center of something that no longer has a single center is a losing game. You can change the platform as often as you like. The center will still not hold.

Here is the shift this series is about.:

In the model I will build across these newsletters, the LMS is no longer the center of gravity. It becomes one tool among many, and not a particularly important one. What matters is the layer beneath it: clean, connected data about how people perform, what support people need, what they did, and what changed, and a set of AI agents that act on that data to do the actual work of learning and development.

Think about everything the current model does through slow, manual, tool-bound effort. Capturing a new capability need from the business. Turning it into clear and measurable performance objectives. Deciding whether learning is the right response. Define clear and measurable learning objectives connected to performance objectives. Decide to build, buy, reuse or do something else entirely. Designing the solution. Checking if it is generating the evidence to know whether it worked. Watching the early signals once it launches. Recommending changes. Recommending where to market it. Deciding what in the catalogue to refresh or retire.

None of that is administration. All of it is judgment and design work that today gets squeezed because we are spending way too much time and effort keeping our software and administration running. In an agentic model, a set of specialized agents carries that work, running on a shared foundation of data, and the human effort moves to where it actually belongs.

That is a big claim, and I am not going to try to prove all of it today. Over this series I will lay it out piece by piece: why the current model is breaking, why AI changes the rules, what foundation the new model rests on, and how to make the transition without needing a multi-year platform program.

But it starts here, with an honest admission that most of us have been avoiding. The reason nobody is happy with the LMS is not that we picked the wrong one. It is that we have been asking the wrong thing of it for twenty years.

You do not need to wait for the rest of the series to begin seeing this for yourself. Three things you can do this week.

First, ask each of the four groups one honest question and just listen. Ask a learner, an administrator, a fellow L&D leader and a business stakeholder what frustrates them about how learning works today. Do not defend the system. Just collect what you hear.

Second, count. For one week, keep a simple tally of how many times you went to an AI assistant, a search engine or a colleague to answer a work question, instead of going to your LMS or LXP. The number will tell you where learning actually happens now.

Third, write down what you genuinely want the LMS to do that it does not. Then look at your list and mark how much of it is administration, and how much of it is understanding, judgment and enabling performance. That line down the middle of your list is the subject of this entire series.

In the next newsletter, I want to go back further, to a dream that shaped a decade of L&D technology decisions and quietly set us up for the frustration we feel today. The dream of standardization, and why the right instinct was aimed at the wrong layer.

Peter Meerman

SLT Consulting — Learning Analytics Made Easy