#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 graduate programme that appeared to succeed
Imagine an organisation celebrating the completion of its annual graduate programme.
Twenty-four graduates have attended the required workshops. Each has completed three rotations, received access to an online learning platform and participated in an innovation challenge. Programme satisfaction is high. Attendance is excellent. The closing presentations are polished, and senior leaders are impressed by the graduates’ confidence.
The programme appears successful.
A few months later, one of the graduates is assigned a moderately complex client problem. She produces a professional-looking analysis within hours. The document is well structured, the recommendations sound credible and the language reflects the organisation’s preferred style.
Her manager initially assumes that the graduate has developed faster than expected.
Then the questions begin.
The graduate cannot explain why one source was trusted over another. She does not recognise that the recommendation conflicts with an earlier client commitment. She cannot distinguish the conclusions she reached from those generated by the AI system she used. When asked what could make the proposed solution fail, she returns to the system for another answer.
She has learned to produce an impressive output.
The organisation does not yet know whether she has developed the capability to understand, question and assume responsibility for it.
This is not a failure of the graduate. She used the tools made available to her and met the visible requirements of the programme.
It is a design problem.
The programme measured participation, exposure and completion. The workplace rewarded the quality and speed of the immediate output. Neither established whether the graduate could perform when the situation changed, the evidence became unreliable or the AI recommendation had to be challenged.
That distinction creates the central question for this essay:
What would organisations design differently if the purpose of early-career development were not programme completion, but readiness for progressively greater responsibility?
When better performance does not mean greater capability
Organisations have traditionally assumed that exposure to work, training, experienced colleagues and progressively challenging assignments will eventually produce capable professionals.
This assumption was never entirely reliable.
An internship can provide exposure without meaningful learning. A rotation can introduce employees to several departments without helping them understand how the organisation functions. Training can improve knowledge without changing workplace performance.
AI makes these weaknesses more difficult to recognise.
Employees can now produce sophisticated analyses, recommendations and presentations with considerably less experience. Their productivity may increase long before their professional judgement develops.
The distinction matters.
Across four experiments involving 3,562 participants, Wu and colleagues found that generative AI collaboration improved performance on tasks where AI assistance was available. However, those improvements did not translate into better performance on subsequent tasks completed without AI.
The research does not establish that AI prevents learning. It demonstrates something equally important: improved AI-assisted performance should not automatically be interpreted as improved human capability.
For organisations, this creates a significant developmental risk.
When managers assess employees primarily through the quality of completed work, they may overlook weaknesses in understanding, critical thinking and independent judgement.
Employees themselves may develop confidence in their performance without recognising how dependent that performance has become on technological assistance.
The challenge is therefore not to prevent inexperienced employees from using AI.
It is to ensure that the technology helps them become more capable rather than merely making their work appear more sophisticated.
Rebuilding development around meaningful work
The idea of apprenticeship provides a useful starting point.
Apprenticeship is commonly associated with trades and technical occupations, but its underlying developmental principle applies much more broadly.
Professionals develop capability by observing experienced practitioners, participating in authentic work, receiving feedback, reflecting on outcomes, and gradually assuming greater responsibility.
The International Labour Organisation’s Quality Apprenticeships Recommendation emphasises structured workplace learning, qualified support, and the development of recognised competencies. Although it addresses formal apprenticeship systems, these principles offer valuable lessons for organisational development programmes.
The new apprenticeship does not require organisations to abandon graduate programmes, internships or job rotations.
It requires them to reconsider what these activities are intended to achieve.
Instead of asking what activities participants must complete, organisations should begin by asking what capabilities they must develop.
Consider a graduate who is expected to become a project manager.
Attending project management training and completing departmental rotations may provide valuable exposure. But readiness requires more.
The graduate must learn to recognise risks, understand stakeholder expectations, evaluate competing priorities and make decisions when information is incomplete.
Those capabilities develop through meaningful participation in work, not simply through observing it.
A developmental pathway should therefore connect three elements:
Purposeful experience: Assignments that expose employees to the situations, decisions and responsibilities they will eventually encounter.
Human guidance: Access to managers, mentors, coaches and experienced professionals who help participants understand the reasoning behind decisions.
Progressive responsibility: Opportunities to demonstrate growing capability before being entrusted with increasingly consequential work.
Rotations, for example, should do more than move employees between departments. They should help participants understand how decisions in one part of the organisation affect another.
Internships should involve authentic contribution within appropriate boundaries, rather than becoming extended observation or inexpensive labour.
The objective is not simply to accelerate employment readiness.
It is to develop professionals who understand their work, recognise its consequences and can contribute responsibly as their careers progress.
AI should strengthen learning, not conceal it
AI creates opportunities to improve workplace development.
It can explain unfamiliar concepts, simulate difficult situations, provide immediate feedback and allow employees to practise decisions before encountering them in real circumstances.
A graduate preparing for a challenging client discussion, for example, could use AI to rehearse questions, explore alternative responses and identify weaknesses in an argument.
These applications can make learning more accessible and provide opportunities that traditional development programmes struggle to deliver consistently.
But AI should not replace the experiences through which professional judgement develops.
A simulation cannot fully reproduce the consequences of a real decision. An AI-generated explanation does not automatically create understanding. Feedback from an AI system cannot always account for organisational context, professional standards or human relationships.
The purpose of AI-supported development should be to strengthen the learner’s ability to think and act, not to remove the need to do so.
This also changes the role of experienced professionals.
A senior consultant may immediately recognise why a proposed recommendation will fail. An experienced manager may notice a stakeholder concern that a graduate overlooks.
For the developing professional, the most valuable learning may come from understanding how that judgement was reached.
Managers and experts therefore need to make their reasoning more visible.
Why was one option rejected? Which assumptions were questioned? What contextual knowledge influenced the decision? Where did AI contribute, and where was its recommendation challenged?
These conversations help employees develop the judgement that cannot be acquired simply by copying a successful output.
Mentoring and coaching also have important, complementary roles.
Mentors can provide professional perspective and help employees interpret organisational realities. Coaches can support reflection, self-awareness, behavioural development and personal responsibility.
AI can strengthen these developmental relationships, but it should not become a reason to withdraw human guidance.
Most importantly, employees must learn to distinguish between what AI contributed and what they themselves understand.
The ability to produce an answer is useful.
The ability to explain, question and take responsibility for that answer is what develops professional capability.
Measure readiness, not programme completion
If organisations want development programmes to produce future capability, they must change how success is evaluated.
Traditional measures such as attendance, programme completion, participant satisfaction and retention remain useful.
But they cannot establish whether employees are becoming ready for greater responsibility.
A graduate may complete every requirement and still struggle to evaluate evidence, challenge an incorrect recommendation or recognise the consequences of a decision.
The new apprenticeship should therefore assess capability through increasingly demanding work.
Participants should have opportunities to observe, practise, explain their reasoning, contribute under supervision, reflect on outcomes and demonstrate that their learning transfers to unfamiliar situations.
As capability becomes evident, responsibility can increase.
Assessment should also distinguish between performance with AI assistance and the underlying human capabilities necessary to use that assistance responsibly.
An employee might produce an analysis using AI but still be expected to defend its assumptions, identify unreliable evidence and explain the limitations of the recommendation.
In some circumstances, demonstrating essential understanding without AI assistance may also be appropriate.
The focus should remain on credible evidence rather than the appearance of competence.
The OECD’s work on skills-first approaches similarly emphasises the importance of credible skills validation, effective learning systems and fairness in assessing capability.
For organisational leaders, four questions become particularly important:
What future capability are we developing? Are our programmes aligned with the specialist, managerial and leadership capabilities the organisation will need?
What experiences will produce that capability? Are participants undertaking meaningful work, receiving feedback and learning from experienced professionals?
What might AI be concealing? Can participants explain their reasoning, evaluate evidence and recognise when AI-generated recommendations are inappropriate?
How will we know when someone is ready? What must participants demonstrate before assuming greater responsibility?
These questions shift the conversation from development activity to developmental outcomes.
Ultimately, the success of a graduate programme should not be determined only by how many people complete it.
It should be determined by whether those people become more capable of making sound judgements, contributing meaningfully and accepting responsibility for increasingly complex work.
A different leadership conversation
The new apprenticeship is not a call to recreate the past.
Traditional career development was often slow, exclusive and dependent on chance access to influential people. Some junior work offered little developmental value. Time served could be mistaken for competence. Mentoring and coaching were not equally available, and many capable people never received the opportunities through which their potential could become visible.
AI creates an opportunity to improve this system.
It can widen access to explanation, allow repeated practice, create varied simulations and provide feedback at a scale that human experts alone cannot match. It can help employees contribute meaningfully earlier in their careers and give managers richer ways to support development.
That opportunity will be lost if AI is used only to make inexperienced employees produce experienced-looking work.
The new apprenticeship must connect technological assistance with authentic experience, human relationships, reflection, verification and progressively greater accountability.
Graduates should learn more than how to prompt a system.
They should learn how to frame problems before asking for answers, test evidence before trusting conclusions, recognise where context changes the recommendation and accept responsibility for the decisions they help shape.
Managers should learn more than how to monitor completion.
They should learn how to make reasoning visible, create developmental assignments, assess readiness and preserve the human encounters through which professional judgement grows.
Organisations should measure more than whether people finish the programme.
They should determine whether potential is becoming capability and whether capability is becoming responsible contribution.
THINK FUTURE Principle
The new apprenticeship should not teach people merely to complete work with AI. It should develop their ability to understand, question and assume responsibility for AI-enabled work.
The new apprenticeship will not be defined by whether graduates attend training, complete rotations, or gain access to AI.
It will be defined by what they can responsibly do as a result.
The final leadership question is therefore:
If your organisation redesigned development around capability rather than programme completion, what would people have to demonstrate before being trusted with greater responsibility?
Coming next
What If AI Makes Human Judgement More Valuable?
The new apprenticeship must prepare people for a world in which information, analysis and plausible recommendations are increasingly abundant.
That does not necessarily make human contribution less important. It changes where its value may lie.
When AI can produce answers in seconds, organisations need people who can frame the right problem, validate what can be trusted, navigate complexity, integrate knowledge and exercise accountable judgement.
In the next essay, we examine a counterintuitive possibility:
What if increasingly capable AI makes well-developed human judgement more valuable rather than less?