Is AI Dismantling the Skills Ladder
There is an old framework for understanding how expertise is built. It predates the modern workplace by centuries, but it maps onto professional development with precision. And right now, AI is breaking it in ways most organisations have not even begun to notice.
“Apprentices spend months just learning how to wring out hot towels. After that, they are taught how to prepare the fish. And then, finally, they are allowed to cook the rice. It is not an easy path.”
The framework is this: mastery develops in three stages. The Apprentice follows instructions - recipes and algorithms, the explicit steps that produce a known result. The Journeyman has internalised enough of those recipes to begin modelling behaviour - adapting, blending, responding to context with a working knowledge of what good looks like. The Master operates from principles. They are no longer following a map or copying a pattern. They are navigating from a deep, internalised understanding of why things work the way they do.
Most professional careers - in finance, engineering, law, medicine, management, design - follow exactly this arc. You spend your early years in structured execution. You progress into contextual judgement. If you develop far enough, you reach a place where you can operate confidently in genuinely novel situations, because you understand the underlying structure of the problem.
That pipeline is now under serious threat.
AI as Journeyman
Large language models are extraordinarily capable within a specific band. They have been trained on vast amounts of human-generated content - the accumulated output of journeymen across every discipline. They can model behaviour, synthesise patterns, produce outputs that look like the work of a competent professional.
But they cannot currently rise above that ceiling. A language model does not hold principles; it holds statistical relationships. It cannot reason from first principles in a truly novel context, because it has no principles - only the reflected surface of other people’s judgement. When a situation falls genuinely outside the distribution of its training, it either fails silently or bluffs with apparent confidence.
This is not a criticism of the technology. It is a description of what the technology fundamentally is. The problem is what happens when organisations treat journeyman-level capability as a substitute for the full skills ladder, as they are doing today.
The Apprentice
The most immediate damage is happening at the bottom of the ladder. Junior professionals today have access to tools that can produce a credible first draft of almost anything - a strategy memo, a financial model, a legal clause, a piece of code. On the surface, this looks like productivity.
The apprentice phase is not just about producing outputs. It is about building the cognitive architecture that makes the journeyman phase possible. Struggling with a problem that is just beyond your current capability, making mistakes, receiving feedback, iterating - this is not inefficiency. It is the mechanism by which expertise is actually constructed.
When an apprentice delegates that struggle to an AI, they get the output but skip the learning. The recipe gets executed without being understood. The rung on the ladder is bypassed, not climbed.
The Master
The longer-term consequence is less visible but more serious. Masters do not appear from nowhere: they are drawn from the pool of journeymen who have developed sufficiently to begin working at the level of principles.
If AI is interrupting the arc from apprentice to master, the journeyman cohort that emerges is shallower. It has the surface characteristics of competence without the deep roots. And from a smaller journeyman cohort, you get fewer masters.
In the past, shifts in the workplaces played out over years and decades. But organisations that are scaling quickly with AI-augmented junior teams are making a fast bet that they will not need master-level expertise in the future - or that they can import it from somewhere else when they do. Both assumptions are fragile, highly risky and highly disruptive.
The Irony
Here is the part that deserves more than a passing nod. The new pain of many workers today is the time it takes to herd and constrain the AI tools and agents we are using. We’re writing prompts, imposing rigid structures, layering in guardrails…all to engineer more predictable, deterministic outputs.
In other words, we’re trying to make the journeyman behave like an apprentice who needs to follow a recipe.
The very capability that makes LLMs interesting - their ability to model complex behaviour across varied contexts - is being deliberately suppressed by users who are uncomfortable with the ambiguity that capability produces. The attempt to constrain the model is an attempt to reduce it to the thing it was never designed to be: a reliable algorithm.
This tells us something important about where many organisations currently sit on the skills ladder themselves. If your response to a powerful tool is to try to reduce it to a simple machine, that is an apprentice-level response to a journeyman-level technology. Are we getting value from the cumulative effort and cost of this technology?
The Practical Implications
In skills development: deliberate friction is a feature, not a bug. If you are managing a team, the question is not how to remove the difficult parts of learning through AI-assisted shortcuts. The question is how to preserve them. Where is the productive struggle in your team’s workflow? Where are people being stretched just beyond their current competence? If AI is eliminating that, you are running a productivity machine that will atrophy your organisation.
In mentorship: the journeyman-to-master transition is not taught. Senior people in your organisation need to be visible in their reasoning, not just in their outputs. When a master works through a problem, the value is not in the answer - it is in watching how they hold uncertainty, weigh competing principles, and make judgements that cannot be fully explained. If you are not creating conditions for that kind of exposure, you are relying on osmosis and hoping for the best.
In organisational design: there is a genuine economic case for protecting master-level expertise. Masters are not simply very experienced journeymen. They are qualitatively different - capable of handling genuinely ambiguous situations where journeyman-level pattern matching breaks down. In stable, well-defined domains, you can get by without them for a long time. In volatile, novel, or high-stakes environments, they are the only thing standing between competence and catastrophe. The economic value of that is invisible, until it is suddenly not.
The Question for You
For those running organisations, the core challenge is this: how do you create an environment that preserves the conditions for mastery, while also capturing the genuine productivity benefits of AI?
The answer is unlikely to be found in policy. Rules about when AI can and cannot be used will be gamed, resented, or simply ignored. The more durable answer is cultural and structural.
It requires leaders who can articulate what mastery actually looks like in their domain - not in terms of outputs, but in terms of judgement. It requires mentorship structures that make the reasoning of experienced practitioners visible and discussable. It requires performance frameworks that distinguish between output quality and developmental progress. And it requires a willingness to accept short-term inefficiency in exchange for long-term capability depth.
Without principles guiding progression up the ladder, organisations are optimising for today’s throughput at the cost of tomorrow’s judgement. At precisely the moment when the business environment is becoming more volatile, more ambiguous, and more in need of principled leadership.
(This is my original thinking, developed and produced with AI support from Claude Sonnet 4.6.)