#THINKFUTURE
Series: Building the Future on a Cracked Foundation
About this series
Artificial intelligence has become one of the defining leadership topics of our time. Organisations are investing billions in technologies that promise greater productivity, lower costs and faster decision-making. Those investments matter, but this series begins with a different question:
What if the greatest challenge of AI is not how it changes work, but how it changes the way organisations develop capable people?
In this THINK FUTURE series, Building the Future on a Cracked Foundation, we explore that question through the lenses of research, organisational psychology and practical leadership. Each essay challenges a common assumption and invites leaders to think differently about capability, work and organisational success.
The professional who practised with AI
Imagine a newly appointed consultant preparing for a difficult client meeting.
The client has rejected an earlier recommendation and is likely to challenge the revised proposal. The consultant’s manager is travelling and the next team meeting is two days away. Rather than waiting, the consultant opens an AI assistant and describes the situation.
The AI assumes the role of the client. It challenges the consultant’s evidence, questions the commercial assumptions and raises possible objections. After each response, it provides feedback and suggests a clearer way to explain the recommendation.
The consultant repeats the exercise several times.
By the morning of the meeting, the presentation is more focused and the consultant feels prepared. The manager reviews the slides and is impressed by the improvement.
The AI appears to have done what a capable manager, mentor, or coach might once have done. It created a practice opportunity, asked questions, provided feedback, and helped the employee prepare for a demanding professional situation.
Yet something remains uncertain.
The AI does not know that the client’s resistance is partly connected to an earlier project that damaged trust. It cannot observe the consultant’s discomfort when challenged by a senior executive. It does not carry responsibility for the advice, the relationship, or the consequences of getting the decision wrong.
It helped the consultant practise. Whether it helped the consultant develop professional judgement will depend partly on what happens next.
Does the manager ask how the consultant reached the recommendation? Does an experienced colleague explain the history behind the client’s resistance? Does the employee receive feedback after the meeting? Is the consultant helped to reflect on what worked, what failed and what should change?
Or does the organisation conclude that the AI has already taken care of the development?
This creates a new leadership question:
When AI can explain, question, simulate and provide feedback, what still requires a manager, mentor, coach or professional community?
The hidden weakness in development through work
For decades, organisations have relied on a simple assumption: put less-experienced people alongside experienced colleagues, give them work and professional capability will emerge.
Sometimes it does. But workplace development has never been automatic. Managers differ in the attention they give to learning. Access to mentors is uneven. Teams transmit valuable professional habits, but they can also pass on poor practices and unquestioned assumptions.
AI changes this arrangement because it can now provide some of the interactions previously supplied by people: explanations, practice scenarios, questions and immediate feedback. That is an opportunity. Employees need not wait for a manager to become available before exploring a concept or rehearsing a difficult conversation.
The risk is more subtle: organisations may confuse access to developmental support with development itself. A helpful response is not evidence that someone has acquired professional judgement, and a polished output does not tell a manager how independently the employee can reason.
Tomorrow’s experts will not develop through a single teacher. They will develop within a system in which people, experience and AI contribute different things.
AI can teach. Real work teaches something else.
AI is particularly useful for explanations, worked examples, repeated practice and structured feedback. In a randomised trial involving 194 undergraduate physics students, a purpose-built AI tutor produced stronger short-term learning gains in less time than the comparison classroom lesson. The tutor was deliberately designed around learning principles; the study did not establish that a general chatbot could replace workplace learning or professional mentoring.
A consultant can rehearse a difficult client conversation with AI. That rehearsal may improve preparation. It cannot fully reproduce the ambiguity, relationships and consequences of the actual meeting.
Experience teaches through attempting work, seeing what happens, receiving credible feedback and adjusting future judgement. A senior colleague may recognise why a seemingly reasonable recommendation could damage a relationship. A manager may know which organisational constraints matter. The employee must eventually make decisions and take responsibility for them.
The question is not whether simulation or real experience is better. It is how one prepares a person for the other.
What managers, mentors, coaches and teams still contribute
AI can resemble several developmental roles in conversation. It can explain like a teacher, advise like a mentor and ask questions that sound like coaching. Similar language, however, does not make these contributions interchangeable.
Managers: connecting learning to performance and readiness
Managers shape the work through which people develop. They assign responsibilities, explain standards, observe performance and decide when someone is ready for a more demanding task. Research on workplace learning highlights the importance of work design, feedback, autonomy and social support.
AI can reduce repetitive explanations and help prepare practice activities. But the manager still needs to ask how the employee reached a conclusion, what they understood and where they need experience. Reviewing only an AI-polished deliverable risks losing sight of the thinking behind it.
Mentors: providing context, identity and opportunity
Mentoring contributes more than advice. Experienced professionals explain unwritten expectations, share lessons from consequential decisions, introduce networks and sometimes advocate for career opportunities. Meta-analytic research has associated mentoring with favourable career outcomes, while also showing variation in its effects.
AI can help an employee prepare questions or explore career options. It cannot independently build a reciprocal professional relationship or personally sponsor someone’s progression.
Coaches: strengthening reflection, agency and accountability
Coaching is not simply a sequence of questions. It is a structured developmental relationship that helps a person examine assumptions, clarify choices and take responsibility for action. Research has found positive effects of workplace coaching on learning and performance across the studies reviewed.
AI can support routine reflection and goal tracking between conversations. Some research suggests people can perceive a positive working alliance with AI-based coaching, but that does not establish equivalence with human coaching. A professional coach brings judgement about the person, their context, ethical boundaries and the relationship itself. This matters particularly when development involves identity, difficult behaviour, relationships or significant consequences.
Teams: developing shared judgement
Much professional learning happens through questions asked of colleagues, explanations of why a decision was made and conversations after something went wrong. These exchanges also develop trust and collective standards.
When employees ask AI instead of a colleague, the answer may come faster, but the developmental need may become less visible to the team. Used differently, AI can help teams prepare after-action reviews, compare perspectives and test assumptions together.
AI should strengthen these relationships, not quietly replace the learning that happens through them.
The apprenticeship risk: when good work conceals weak learning
In a professional apprenticeship, novices observe experienced practitioners, attempt work, receive feedback and gradually assume greater responsibility. Cognitive apprenticeship makes expert thinking visible: how a problem is framed, which evidence matters and why a decision is justified.
AI can help model reasoning and create progressively harder exercises. But it can also complete the reasoning before the employee attempts it. If support never decreases, the organisation may mistake AI-assisted performance for independent capability.
Employees therefore need two kinds of literacy: knowledge of their profession and an understanding of AI’s capabilities and limitations. Neither alone is enough. A capable AI user without domain knowledge may accept a persuasive error; a domain expert who misunderstands AI may overlook useful assistance or rely on it in the wrong circumstances.
Learning should include imperfect AI answers, missing evidence and occasions when the employee must explain why the system’s recommendation should be challenged or rejected.
Designing a better teaching system
The leadership response is not to remove AI from professional development. It is to make the responsibilities within that development explicit.
Use AI for access and practice. Offer explanations, simulations and structured feedback, with clear boundaries about where those outputs can be trusted.
Keep managers accountable for readiness. Ask employees to show their reasoning, connect assignments to developmental goals and review what was learned after consequential work.
Protect human mentoring and coaching. Reserve time for professional context, relationships, sponsorship, difficult reflection and developmental accountability that a tool cannot independently assume.
Make expert thinking visible. Create opportunities for experienced professionals to explain choices and uncertainties rather than simply sharing completed outputs.
Test independence and govern use. Gradually reduce AI support, verify performance in authentic situations and set clear rules for confidentiality, evidence, professional standards and escalation.
The organisation must also decide what constitutes adequate evidence of growth. Faster task completion or better-looking work may be valuable performance outcomes, but neither, by itself, proves that judgement has developed.
UNESCO’s human-centred guidance for generative AI in education reinforces the importance of agency, privacy, deliberate learning design and institutional responsibility. Workplace conditions differ, but the underlying question remains: who owns the quality and consequences of the learning environment?
AI can extend access to teaching. The organisation remains responsible for developing its people.
A different leadership conversation
Who is teaching tomorrow’s experts?
The answer is no longer simply the manager, mentor, coach or experienced colleague.
AI is becoming part of the teaching environment. It can make explanations available on demand, provide repeated practice, simulate difficult situations and extend feedback beyond the limited time of human experts. Used deliberately, it may help more people develop more quickly.
That does not make the human contributors obsolete.
Experience still connects decisions to consequences. Managers still shape work and judge readiness. Mentors still transmit professional context and open access to communities. Coaches still help people examine their thinking and take responsibility for change. Teams still develop shared practices. Professional communities still protect standards that no single employer or technology should redefine.
AI can strengthen each of these relationships. It can also displace them if leaders mistake accessible interaction for complete development.
The organisation must decide what the system is allowed to teach, which knowledge must be verified, where human judgement is indispensable and who remains accountable for the person’s progression.
The future of expertise will therefore not be determined by whether AI can teach.
It will be determined by whether organisations can combine technological reach with human context, relationships and responsibility.
THINK FUTURE Principle
AI can extend access to explanation, practice and feedback. Developing responsible expertise still requires human context, relationships and accountability.
If AI can perform part of the teaching, the leadership question is no longer simply who provides the information.
It is:
When AI can teach part of the work, who remains responsible for developing the professional?
Coming next
The New Apprenticeship
Knowing that tomorrow’s experts require a broader teaching system is only the beginning.
Organisations must translate that insight into the design of graduate programmes, internships, rotations, coaching, human–AI work and capability development. They must decide which experiences to preserve, which tasks to redesign and how responsibility should increase as capability develops.
In the next essay, we move from the question of who teaches to the practical architecture of development:
If tomorrow’s experts need a new teaching system, what should the new apprenticeship look like?