The Architecture of Leadership
Exploring the systems that shape the practice of leadership
The Return of Apprenticeship in the Age of AI
Access to output is expanding, but the path to judgment is becoming less visible
There is a growing belief that access to artificial intelligence is flattening the learning curve. Tasks that once required years of experience can now be assisted, accelerated, or partially automated. Drafts can be generated, analyses can be summarized, and decisions can be supported in ways that feel immediate. From the outside, this creates the impression that capability is becoming more evenly distributed across experience levels.
What is less visible is what is being removed from the process that used to build that capability. Much of professional development has historically come from time spent in proximity to the work before full ownership was expected. Early-career employees observed how decisions were made, sat in conversations they did not fully understand, and learned through repetition, correction, and exposure. That process was not always efficient, but it created a pathway for developing judgment over time.
That pathway is becoming less defined as AI becomes more embedded in how work is completed. Instead of working through ambiguity, individuals can generate outputs quickly. Instead of drafting from scratch, they can refine something that already exists. Instead of sitting with uncertainty, they can move immediately to structured responses. While this accelerates productivity, it changes how understanding is formed and what parts of the process individuals actually experience.
The risk is not that people become less capable in an absolute sense. The risk is that the process through which capability develops becomes less consistent and harder to observe. When outputs improve faster than understanding, it becomes difficult to assess where someone is in their development. The signal of capability begins to separate from the appearance of capability.
This dynamic already exists in many organizations, even without AI. There has always been tension between speed and development. Leaders are asked to deliver results, move quickly, and operate efficiently, while development requires time, repetition, and space for imperfection. Apprenticeship, in its traditional form, allowed individuals to see work before they were responsible for it and to engage with it before they were expected to master it. These conditions are often the first to be compressed when speed becomes the priority.
AI intensifies that compression by making it easier to bypass the parts of the process that create learning. A junior employee can now produce something that appears polished with assistance, but may not fully understand the reasoning behind it. The result improves, but the development of judgment may not keep pace. Over time, this creates a gap that is not immediately visible but becomes more pronounced as responsibilities increase.
This showed up in an organization where early-career employees were contributing to strategic discussions. The quality of their materials had improved significantly, and the work appeared thoughtful and complete. When asked to explain how conclusions were reached or how tradeoffs were considered, the depth of understanding varied widely. The product created confidence, but the underlying reasoning was not always as developed as it appeared.
A similar pattern began to emerge in how work was assigned over time. Individuals who could produce polished outputs quickly were given more responsibility earlier. From the leadership perspective, this made sense. The work appeared strong, and the individual seemed capable. But as expectations increased, the gap between output and underlying judgment became more visible. When faced with situations that required independent decision-making without clear inputs, the same individuals struggled to navigate tradeoffs that had previously been handled by the tools or by others.
This felt like a sudden and steep drop in performance. Yet with a bit of distance, it reflected a different issue: The development pathway had been shortened in ways that removed exposure to the parts of the work that actually build judgment. What appeared to be acceleration was, in some cases, a reordering of the learning process rather than a replacement for it.
Situations like this are not the result of individual shortcomings. They are the result of how work is structured and how tools are integrated into that structure. When individuals are consistently positioned to refine rather than originate, or to respond rather than wrestle with ambiguity, the opportunities to build judgment become more limited. The development process becomes narrower even as productivity increases.
This is where the idea of apprenticeship becomes relevant again, not as a return to the past, but as a response to what is being lost. Apprenticeship is not defined by hierarchy or formality. It is defined by exposure, repetition, and proximity to decision-making. It creates conditions where individuals can see how work is done, not just what the final product looks like, and where they can build understanding through participation rather than assumption.
Reintroducing these conditions does not mean slowing everything down or limiting the use of AI. It requires being intentional about how work is experienced. Leaders may ask individuals to walk through their reasoning, not just present conclusions. They may create space for early-stage thinking before refinement occurs. They may involve less experienced employees in conversations where they are not yet expected to contribute fully but can observe how decisions are made.
This also requires clarity around what is being evaluated. If outputs are the only measure, then tools will naturally optimize toward that specific outcome. If understanding and judgment are also valued, then the process behind the product needs to be visible. Asking how a decision was reached, what alternatives were considered, and where uncertainty still exists provides insight into development that the final deliverable cannot.
There is also a responsibility on the individual side that becomes more pronounced in this environment. Access to tools can create the impression that understanding is optional because the work meets expectations. Over time, that gap becomes more difficult to sustain. When individuals are asked to operate without assistance or in situations where the tool does not provide clear direction, the absence of underlying judgment becomes more visible.
This is where expectations need to be more explicit than they have been in the past. Early-career employees still need to spend time understanding how work actually happens, not just how it is produced. This includes observing how decisions are made, building relationships that influence outcomes, and engaging with ambiguity rather than bypassing it. The presence of AI does not remove the need to do this work. It increases the importance of being intentional about it.
Organizations that are beginning to recognize this shift are adjusting how they approach development. They are creating space for individuals to explain their thinking, not just deliver results. They are pairing less experienced employees with more experienced ones in ways that allow for observation and dialogue. They are making room for work that is incomplete so that learning can occur before refinement.
These changes do not remove the tension between efficiency and development. The pressure to move quickly will continue, and AI will continue to expand what is possible in shorter periods of time. The tradeoff is not whether to use these tools, but how to ensure that the process of learning is not fully replaced by the process of producing.
This is not an immediate problem. It is something that accumulates over time. In the short term, productivity increases and work appears more polished. Over time, gaps in judgment begin to surface in more complex situations. Decisions take longer when assistance is not available. Escalations increase because context is missing. Leaders spend more time reviewing work that appears complete but lacks depth.
At that point, the absence of apprenticeship becomes visible in ways that are difficult to ignore. Reintroducing it later is more difficult because the habits of working through tools rather than through understanding are already established. It requires recreating conditions for learning within systems that have been optimized for speed.
This is where leadership becomes a design problem. If apprenticeship is not intentionally built into how work happens, it will not emerge on its own. The presence of AI does not eliminate the need for development. It increases the importance of being deliberate about how development occurs and where learning is preserved within the flow of work.
The return of apprenticeship is not about rejecting technology. It is about recognizing that access to tools does not replace the process through which people learn to think, decide, and lead. That process still requires exposure, repetition, and engagement with the work itself over time.
Questions Worth Sitting With
- Where might strong outputs be masking gaps in understanding rather than reflecting true capability?
- How clearly can the organization distinguish between what someone produced and what they actually understand?
- What exposure to decision-making and context is being removed in the name of efficiency?
- Where should speed be preserved, and where should learning be protected, even if it slows the work down?