AI is increasingly performing the routine cognitive work through which people traditionally developed professional expertise. This creates an important human performance question: if technology removes part of the practice through which judgement was built, how will expertise develop in the future?
AI Is Changing More Than the Work People Do
Much of the discussion about AI and entry-level work focuses understandably on employment. As generative AI becomes capable of research, documentation, preliminary analysis and basic coding, organizations can accomplish more with fewer people performing the routine cognitive tasks traditionally assigned to junior employees.
There is another consequence that may prove equally important. Those tasks were never merely productive work. They were also practice. Through repetition, correction, observation and gradually increasing complexity, inexperienced employees developed the knowledge and pattern recognition that eventually allowed them to exercise professional judgement. Removing the task can therefore remove part of the developmental experience embedded within it.
A recent McKinsey article, Building Expertise in the Age of AI: Who Trains the Next Generation?, describes this problem directly, arguing that AI is absorbing precisely the activities through which younger employees historically developed instincts and judgement. It also cites Microsoft engineering leaders Mark Russinovich and Scott Hanselman, whose work describes an “AI boost” for experienced engineers who already possess the expertise to direct and verify AI output, alongside an “AI drag” for early-career engineers who have yet to develop it.
The phenomenon points towards a deeper question for Human Performance Intelligence™: what happens to human capability when technology begins performing part of the work through which that capability was previously developed?
Expertise Is Built Through Exposure
Expertise is sometimes treated as though it were primarily an accumulation of knowledge. Knowledge certainly matters, but professional capability also develops through repeated encounters with situations in which knowledge has to be applied. People begin to recognise patterns, understand exceptions, anticipate consequences and distinguish between situations that appear similar on the surface but require different responses.
This accumulated experience is particularly important for judgement. An experienced professional can often identify that something is wrong before being able to articulate precisely why, because years of exposure have created mental representations against which new situations are evaluated. Junior work historically provided much of that exposure at relatively low levels of risk. The employee performed simpler tasks, received feedback, observed more experienced colleagues and gradually developed the cognitive structures required for more complex work.
AI potentially alters this developmental sequence. If a system generates the first analysis, proposes the solution or produces the initial draft, the employee encounters the task from a different cognitive position. They may still learn from evaluating the output, but evaluation itself requires a foundation against which the output can be judged. The less developed that foundation is, the greater the possibility that plausible AI output will substitute for rather than strengthen independent reasoning.
Performance and Capability Can Begin to Diverge
This connects to an important distinction within Human Performance Intelligence™ between the performance someone can produce within a particular system and the underlying human capability contributing to that performance. AI can increase the former immediately because the combined human-AI system has access to capabilities that the individual alone does not possess. Whether those gains become part of the individual’s own capability depends on what happens during the interaction.
For experienced professionals, AI can function as an amplifier because they already possess sufficiently developed mental models to interrogate its reasoning, recognise weak assumptions and integrate useful outputs into a broader understanding of the problem. For novices, the relationship is more complicated. The technology can allow them to produce work that appears considerably more sophisticated than their current expertise would independently support, potentially narrowing the visible performance gap between novice and expert without equivalently narrowing the underlying capability gap.
This makes observable performance increasingly difficult to interpret. A strong output tells us something about the effectiveness of the human-AI system, but considerably less than it once did about the expertise of the individual within it.
AI Could Also Accelerate Expertise
None of this implies that AI must weaken human development. The same technology that removes traditional forms of practice can potentially create better ones. AI can provide immediate feedback, expose employees to alternative reasoning, simulate difficult situations and allow people to encounter a much broader range of problems than conventional apprenticeship made possible.
The critical variable is how the interaction is structured. An employee who attempts a problem before comparing their reasoning with an AI-generated solution is engaged in a different cognitive process from someone who receives the solution first and edits it. An employee required to explain why an AI recommendation is appropriate is practising different capabilities from someone rewarded primarily for producing the correct output quickly. Small differences in work design can therefore determine whether AI primarily substitutes for cognitive effort or becomes part of the mechanism through which capability develops.
From an HPI perspective, this is where the relationship between Cognitive Load and Adaptive Capacity becomes particularly important. Development requires enough challenge to stimulate learning without creating demands that exceed an individual’s available capacity. AI gives organizations an unprecedented ability to alter that balance, removing unnecessary cognitive burden while potentially preserving, or even deliberately creating, the forms of effort through which expertise grows.
The Human Performance Question
The arrival of AI therefore creates a more complex question than whether junior employees will still be needed. Organizations will continue to need people capable of exercising judgement, solving unfamiliar problems and recognising when an apparently convincing answer is wrong. The developmental pathways through which people acquire those capabilities, however, can no longer be assumed to remain unchanged.
Human Performance Intelligence™ suggests that organizations should distinguish carefully between work that is inefficient because humans are performing tasks technology can do better and work that appears inefficient because people are still developing capabilities they will need later. Both may look similar on a productivity dashboard, while having very different implications for long-term performance.
As AI becomes more capable, protecting human development does not require preserving obsolete work. It requires understanding what made certain experiences developmental in the first place and designing new human-AI interactions that reproduce, or improve upon, those mechanisms. The future of expertise may depend less on how much work AI can remove than on how intelligently organizations redesign the learning that remains.