Part 2 - What Montessori Understood About Learning That Corporate L&D Forgot

Skills won't save you. Neither will AI (or could it?)

In Part 1, I argued that corporate learning ecosystems are digital pantries (though I do like Shellie Grieve's analogy of a library). Stuffed with content, tagged poorly, while we hope people figure out how to cook. The Montessori method, designed in 1907 for toddlers, still outperforms most of what we build.

Part 2 is about the things we keep reaching for to fix this: skills taxonomies, corporate universities, and AI-powered recommendations. And why are they still the same same, but different.

The Skills and Tasks Distraction

The L&D industry is obsessed with skills. Taxonomies. Ontologies. Skills-based organizations. Before these, it was competencies. Soon it'll be tasks. Every few years, we convince ourselves that a new organizing principle will finally make learning work and be able to get our seat at the table.

We're arguing about skills versus tasks while often ignoring that neither addresses how people actually develop capability (don't get me wrong, a lot of teams are working on it; it's just missing from many narratives).

Montessori doesn't track "skills acquired" like checkboxes. It tracks developmental progression. Where is this child? What are they ready for next? The tracking is continuous, observational, and developmental. Not binary.

Most corporate skills-tracking still operates closer to a binary than a developmental model. Even platforms with proficiency levels tend to get used as checklists. You're a 3, or you're not. The evidence of "what are you ready for next" often gets lost (remember digital portfolios, people).

My kids' teachers can tell you exactly where each child is across multiple domains. Not because they administered a test. Because they observe each child constantly throughout their work cycles. They know which materials each child has mastered, what they're working on, and which they're not ready for. The tracking serves learning, not compliance.

When was the last time your L&D team could articulate where specific employees are in their developmental progression? Not what courses they've completed. What are they ready to learn next as it relates to the initiatives the company needs to execute? What work projects have been added to their professional portfolios showcasing actual competency? What are competency signals showing that they have mastered something and are ready to progress to something more challenging?

Skills taxonomies aren't at all useless. They should be part of the foundation, not the whole kitchen. You need to know where the organization is heading before you map skills. You need a well-designed learning environment before you build content. Buying every ingredient at the farmers market before you know what you're cooking just fills a fridge with stuff that'll go to waste (I am guilty of this because who knows, maybe I do urgently need 12 bundles of black garlic).

Organizations love building corporate universities. Academies, Institutes, Schools of Excellence. Platform selected. Content aggregated. Then nothing changes. The content is still passive: videos, PDFs, SCORM modules, static skills taxonomies bolted onto learning systems. Skills-tagged playlists don't become learning systems just because we renamed them. Disaggregated video-based performance tools.

If you want to understand what building a skills-based organization actually requires (not the buzzword version, the actual work), read Koreen Pagano's Building the Skills-Based Organization: A Blueprint for Transformation. The subject is genuinely hard, which is exactly why duct tape and baling wire aren't going to fix this one.

The Incentive Trap

I have sympathy for L&D teams, even as I criticize the function.

Montessori requires trust in the learner. Patience. Long-term thinking about capability rather than short-term box-checking. Investing in environment design rather than content production.

Corporate L&D rarely operates under these conditions, as the function is rarely aligned with the work output that the business measures.

We're measured on courses produced. Completion rates. Time to deployment. Satisfaction scores that measure whether people had a pleasant time (I love me a happy sheet), not whether they learned anything. So we build content factories optimized for throughput.

This is a systems problem. I've worked with brilliant L&D professionals who know what good learning solutions look like and how to build them, and are completely hamstrung by metrics that reward the wrong outputs.

You can't design a proper kitchen when leadership just wants you to pump out more frozen meals.

Here's the uncomfortable part: proving value doesn't automatically earn you permission to change.

I've worked with L&D teams that run very successful pilots that leadership ignored because the results didn't fit the narrative. Proof points only matter if someone's willing to look at them and own the number. Before you can redesign metrics, you need a stakeholder who cares more about the business outcome than the completion dashboard.

That's not a learning problem. That's a political problem. And pretending otherwise is why most transformation efforts stall.

The path forward isn't just "prove value." It's finding the executive who owns a number you can move together. Revenue. Cost. Risk. Pick one. Partner on it. When that number moves, you've earned something more valuable than a proof point. You've earned credibility! (let’s hope it moves in the right direction)

So What Do We Do?

Start small, prove value, and protect the scope.

One department. One business problem, tied to revenue, cost, or risk. Build a learning system aligned to that outcome. Use learning theory intentionally. Measure capability against the business metric. Use that proof point to earn the right to do more.

But most of us aren't improvising from experience yet. We need scaffolding. A clear framework, a measurable outcome, and permission to fail before the stakes get high. (My wife has endured years of this philosophy applied to actual cooking. She remains patient and honest. Both are underrated virtues.)

But here's what most advice skips: you need a way to protect this work from scope creep.

Without a systematic filter, your "one problem" becomes seventeen problems. We use a version of this triage at 8Levers, but the principle is simple. Before committing resources, some sample questions:

• Is there evidence of a learning need? (Not "someone requested it." Actual evidence.)
• Does it impact risk, safety, compliance, or revenue? (If yes, fast-track. If no, keep asking.)
• Is there stakeholder sponsorship? (Real sponsorship, not a signature on a form.)
• Can the culture support the performance goals? (If not, learning won't fix it.)

The full triage consists of 9 questions, each generating data. That data becomes evidence when stakeholders ask why their pet project got deprioritized. You're not saying no. You're saying, "Here's where this sits in the queue, and here's why."

A resource allocation model that we advocate at 8levers: 30% of L&D capacity goes to non-negotiables (compliance, safety, regulatory). The remaining 70% gets ruthlessly triaged by business impact. Not the arrival order. Not political pressure. Impact (usually performance support resources not courses).

Change what you measure: Report on performance metrics that the business already tracks. Partner (or shadow) with whoever owns the number you're trying to move. Even imperfect correlation beats no correlation.

Reframe L&D's role: Stop being order-takers. When a stakeholder says, "build me a course," push back. What business problem? What does capable look like? How will they know if it worked? If they can't answer, don't build. Triage ruthlessly. If it doesn't connect to making money, saving money, or reducing risk, it goes to the back of the queue.

Design systems, not content: Before building anything: What's the prepared environment? How does knowledge flow? Where does practice happen? How does it happen? What does progression look like? If you can't answer these, you're not ready to build.

You probably can't redesign everything at once. But you can change one project. One team. One onboarding cohort.

The Missing Role

In Montessori, the guide is everything.

Not a lecturer. Not an administrator. Not a coach who may or may not have time to help you. Someone (or something) trained to observe constantly, understand developmental readiness, introduce the right challenge at the right moment, and then step back. It's a skill that takes years to develop. My kids' guides can read subtle signals. When a child is ready for something new, when they need to sit with difficulty a little longer, when frustration is productive rather than shutting them down.

Most corporate L&D functions don't have this role. Some boutique or heavily resourced teams do, which kind of proves the point about what it takes.

We have content builders, LMS admins, and facilitators. Maybe performance consultants or coaches. But nobody whose job is ongoing observation and real-time developmental judgment.

We've designed the guide out of the system (I get it, it is impossible to scale. Then we wonder why the system doesn't develop people. I have ideas about this…

The AI Question

AI could, in theory, help us build something nearly impossible at scale: learning ecosystems that scaffold, adapt, and respond to individual readiness. An AI that functions like a Montessori guide. Observing where someone is, understanding what they're ready for, and introducing the right challenge at the right moment. Feedback loops that update capability data as it streams into performance systems.

Theoretically.

Most organizations will screw this up. Most vendors will market a solution to the problem they have not fully understood. This part is hard. It is also the Utopian promise that L&D has been working towards for 50 years (anyone remember the promises of intelligent tutoring systems from the 1970s, 80s, 90s, and 2000s?).

Remember when every vendor promised us "the Netflix of learning"? Personalized recommendations that would accelerate how employees develop? We got algorithmic content suggestions based on your job title and on what people like you clicked. That's not personalization. That's a recommendation engine with a marketing problem. I haven't seen an AI chatbot yet that does much better. Most are built on a faster, more overconfident approach to the same irrelevant suggestions (with some nice pandering to our egos). The technology is improving faster than the implementations. (Show me somewhere I'm wrong, please.)

AI doesn't fix broken learning systems. It operationalizes the brokenness faster. Deploy it on top of a content repository with no grounding in learning theory, no linkage into business datasets, no feedback loops that update capability data in real time, and you get more content nobody needs at unprecedented speed.

Congratulations. You've automated mediocrity. This is a great place to mass-produce delicious chocolate-covered broccoli Ed Wills.

There's another problem nobody wants to talk about: data fragmentation.

You can't build an AI learning guide without clean skills data, performance data, and learning interaction data flowing into the same place. Most organizations have these in different systems that don't talk to each other (most actually don't even know where this data sits or how to access it). HR owns performance and a skills taxonomy. L&D owns the LMS and built a different skills taxonomy. IT owns the data lake. Finance owns headcount planning.

The AI isn't the blocker. Understanding the plumbing is.

Before you buy any AI learning tool, apply the same triage logic: can you connect learning activity to performance outcomes in a single view? Can you push data from this tool into a centralized reporting stream? If not, you've identified the actual work. The AI can wait.

AI amplifies whatever you point it at. The organizations that benefit will be the ones doing the unglamorous work first: fixing data infrastructure, aligning systems, building feedback loops. Then AI becomes powerful because it's amplifying something that actually works.

My kids' Montessori guides don't need AI. They have designed environments, deep training, and time to observe. AI could give corporate L&D something similar. But only if we know what we're building toward and have the data infrastructure to support it.

Right now, most organizations deploying AI in learning are making the same mistakes faster. That's not transformation. That's acceleration toward the same end point we have always gotten to.

Time for a Redesign?

AI could be the guide we've designed out of the system. Skills could be a foundation instead of a checkbox. Corporate universities could be genuine learning systems instead of rebranded content libraries.

None of that happens until we stop confusing delivery with development. Start with the business problem. Design the environment. Protect the scope. See what happens.

I'm tired of our 4-year L&D hypercycle. Honestly, it feels a lot like crypto. Still waiting for my millions, BTC gods.

Your turn. What are the holes you see here? (Yes, there are many. I didn't want to write a book on this. Yet.)