Leadership in Motion - Issue 4
The Developmental Architecture of Work
In Issue 3 of Leadership in Motion I concluded that the future of leadership will belong to those who use technology well, while deepening the very human capacities that make leadership worth following: judgement, courage, intuition and connection. AI is increasing the speed at which information can be gathered, options can be generated and work can be completed, while the human systems that create alignment, trust and sound judgement are adapting more slowly.
As I have reflected on that argument, another question has continued to surface. When leaders decide which activities should be automated, augmented or redesigned, how much attention are they paying to what people previously learnt by undertaking them?
Work has always served more than one purpose. It produces an outcome for the organisation, while at the same time creating the experiences through which people develop confidence, capability and judgement. Those two purposes have rarely needed to be designed together because, for many years, they evolved together naturally. People learnt through doing the work, observing others, making mistakes, receiving feedback and gradually taking responsibility for decisions of greater consequence.
AI is beginning to change that relationship.
One of the less visible consequences of redesigning work is that organisations are also redesigning the way people learn. The productivity implications are receiving considerable attention. The developmental implications are much less understood.
That may prove to be an important omission.
Leadership Signals
Several strands of recent research approach this question from different directions, and together they suggest that the effect of AI on human capability will depend considerably on how work is designed around it.
The International Labour Organization's review of empirical research into generative AI and work provides a useful starting point. Drawing together evidence from experiments, organisational studies and representative surveys, it finds that productivity gains are real but uneven and that AI is beginning to reshape combinations of tasks and the organisation of work rather than simply replacing isolated activities. Importantly, some of the research it reviews also suggests that less experienced workers can benefit disproportionately from AI assistance in particular settings.
This creates an interesting possibility. AI may allow people to participate in more sophisticated work earlier in their careers and give them access to knowledge that previously depended on years of experience or proximity to more senior colleagues. The developmental effect could therefore be positive, particularly where technology extends someone's ability to engage with a problem rather than removing the need for them to engage with it.
A recent Perspective published in Nature Computational Science considers this possibility through the lens of the democratisation of knowledge work. The authors examine whether generative AI could make sophisticated forms of knowledge work accessible to a broader population, while also recognising that access to and benefits from the technology remain uneven. Their work raises an important question for organisations: whether AI simply makes existing experts more productive or begins to change who can participate meaningfully in work that previously required considerable expertise.
Research published in Scientific Reports offers another perspective and, for me, one of the more interesting signals. In a preregistered experiment involving 269 participants, followed by a separate survey of 270 workers, researchers compared different forms of AI use during knowledge work. People who relied more passively on AI reported lower independent self-efficacy, ownership and task meaningfulness, while participants who first developed their own thinking and then used AI collaboratively largely avoided those effects.
The distinction matters because it suggests that the developmental question may depend less on whether AI is present and more on the nature of human participation around it. Two people may produce similarly impressive outputs with AI while having engaged very differently with the reasoning required to create them.
A systematic review of 627 peer reviewed studies into human and AI collaboration in decision making reinforces this point from another direction. The research identifies different configurations of human and machine involvement depending on the nature of the decision, suggesting that there is unlikely to be one optimal model for human and AI collaboration. Decisions involving interpretation, ambiguity and adaptation create different requirements from those that can be addressed through established rules and clearly defined information.
The workforce implications are already becoming visible. PwC and the World Economic Forum's research into the future of entry level work, drawing on insights from more than 9,000 early career workers across 48 countries, identifies rapid changes in the skills required in AI exposed occupations and raises questions about entry routes, job design and future talent pipelines. The work at the beginning of many careers is changing quickly enough that established assumptions about how people progress towards more senior responsibility deserve renewed attention.
These sources do not yet tell us what AI enabled work will do to professional judgement over the course of a career. The longitudinal evidence simply does not exist at sufficient scale. What they do show is that AI is changing the nature of people's participation in work, and that different forms of participation appear to produce different human outcomes.
That distinction takes us beyond the usual conversation about skills.
The Systemic Tension
For much of my career, organisations have rarely needed to explain how professional judgement develops because it emerged through the progression of work. People began with relatively bounded responsibilities, gradually encountered greater complexity, observed experienced colleagues, defended recommendations, made decisions and learnt from the consequences.
Experience alone has never guaranteed expertise. What mattered was the combination of experience with feedback, reflection, social interaction, increasing responsibility and repeated exposure to situations where the answer was neither obvious nor complete. Over time, those experiences created pattern recognition and a better appreciation of context, trade offs and consequence.
Together, these mechanisms form what I think of as the developmental architecture of work. They are the largely invisible processes through which activity gradually becomes capability.
AI has the potential to strengthen that architecture considerably. It can provide rapid feedback, expose people to alternative perspectives, broaden access to expertise and allow someone earlier in their career to participate in work that might previously have remained beyond their reach. In the right environment, it is entirely plausible that people could develop some forms of capability more quickly than previous generations.
The same technology can create a different outcome where increasingly capable outputs reduce the need for people to articulate assumptions, wrestle with ambiguity, identify missing information or explain why they believe a recommendation is appropriate. An individual may appear capable of producing more sophisticated work while the organisation has less visibility of how deeply they understand the reasoning beneath it.
This creates a tension between two outcomes that can easily be mistaken for one another: assisted performance and independently held capability.
For senior leaders, the distinction is difficult because the immediate evidence may look entirely positive. Work is completed more quickly, quality improves and less experienced colleagues appear able to contribute at a higher level. All of those benefits may be genuine.
The longer term developmental effects will take much longer to become visible.
An organisation can measure how quickly an analysis was completed or how much productivity improved. It is considerably harder to know whether someone is becoming better able to recognise an unstated risk, challenge a plausible but inappropriate recommendation, navigate a difficult stakeholder relationship or make a sound decision when the available evidence remains incomplete.
The leadership question therefore becomes more subtle than whether AI is developing or deskilling people.
It is whether the new experience of work will develop the forms of judgement organisations will need.
The Leadership Implication
The opportunity for leaders is to become much more intentional about the relationship between work design and capability development.
Whenever an activity is redesigned, there is value in asking two questions rather than one. The first is familiar: what improvement in productivity, quality or customer value can AI create? The second receives much less attention: what did people previously learn through undertaking this work, and where will that learning now come from?
Sometimes the answer will be that little of value has been lost. Routine activity should not be preserved simply because earlier generations happened to perform it, and AI may provide a considerably richer learning environment than the work it replaces.
Other activities carry developmental value that is less obvious. Preparing an analysis may teach someone how to distinguish signal from noise. Building a recommendation may expose weaknesses in their assumptions. Presenting a proposal to experienced colleagues may reveal whether they genuinely understand it. Living with the consequences of a decision may calibrate their confidence in a way that no training programme can easily reproduce.
As those activities change, leaders may need to recreate the underlying developmental mechanisms rather than preserve the tasks themselves.
Managers will be particularly important because they sit closest to the point where AI assisted performance becomes organisational judgement. Their role may increasingly involve understanding not only the quality of the answer but also the quality of the thinking behind it. Asking which assumptions were tested, what alternatives were considered, where the individual disagreed with the technology and what evidence might change their conclusion can turn AI enabled work back into a developmental experience.
Leadership teams may also need to become more precise about what they mean by judgement. Analytical judgement, strategic judgement, interpersonal judgement and ethical judgement are related but different capabilities, and each develops through different combinations of knowledge, experience, reflection and consequence. Treating them collectively as "critical thinking" risks simplifying a much more important organisational question.
The research is still developing, and the long term consequences will become clearer as organisations accumulate more experience of AI enabled work. What can already be seen is that the architecture of work is changing quickly enough that relying on capability to develop in the same way it always has feels increasingly difficult to justify.
Work has always been one of an organisation's most powerful learning systems. As leaders redesign work, they are also redesigning that learning system, whether they intend to or not.
Perhaps the leadership opportunity is to make that design intentional.
Reflection Question
As AI changes the way work is performed across your organisation, which experiences are becoming less common, what did people previously learn from them, and where will that learning come from in the future?
Source Log
Erlend Asker
Executive coach and leadership advisor. Former commercial executive. Works with senior leaders navigating organisational transformation.
Leadership keeps moving
The intention of Leadership in Motion is to explore emerging leadership questions that I believe deserve greater attention. If you would like to follow the thinking as it develops, leave your details below and I’ll send you future issues.
