#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 answer nobody challenged
Imagine an executive team considering a significant investment.
The proposal is persuasive. It includes market analysis, financial projections, implementation scenarios and a detailed assessment of the organisation’s competitors. The document is professionally written, visually compelling and supported by what appears to be credible sources.
It took the project team two days to produce work that would previously have required several weeks.
The executives are impressed.
One director asks how the projected customer adoption rate was calculated. The team explains that the AI system generated the estimate from the market information it had analysed.
Another asks whether the proposed operating model has been tested against South African regulatory and infrastructure conditions. The project leader is uncertain.
A third notices that several sources do not say quite what the proposal claims they say. One citation leads to an opinion piece. Another refers to research conducted in a materially different market. A third cannot be found.
The proposal is not necessarily wrong.
The problem is that nobody in the room yet knows whether it is sufficiently right to shape a consequential decision.
The system has provided an answer. The team has refined and presented it. The executives must now decide what deserves to be believed, what requires further investigation and what consequences they are prepared to accept.
The speed with which the answer was produced has not reduced the need for judgement.
It has concentrated it.
This creates the central question for this essay:
What if increasingly capable AI does not make human judgement less valuable, but makes well-developed judgement more consequential?
When answers become abundant
Much of the discussion about artificial intelligence begins with a reasonable assumption.
If machines can perform more cognitive tasks, the value of human contribution must decline.
There is evidence supporting part of this concern.
Generative AI can already perform work previously associated with educated professionals. It can draft documents, analyse information, generate software and propose solutions. Some tasks will be automated, and some roles will change substantially.
The productivity gains are significant.
In an experiment involving 453 college-educated professionals, researchers found that access to ChatGPT reduced average completion time on writing tasks by 40 percent while increasing output quality by 18 percent.
These findings matter. Leaders should not minimise what AI can do merely to preserve a comforting account of human importance.
But improved machine capability does not necessarily mean that human judgement becomes less important.
AI changes the supply of cognitive output.
Information becomes easier to obtain. Analysis becomes faster to produce. Plausible recommendations become cheaper to generate.
The scarcity may therefore move.
Organisations may have more answers while becoming less certain about which answer addresses the right problem, rests on trustworthy information, or deserves to influence action.
Value shifts from producing answers towards determining which answers matter, which can be trusted and what should happen next.
Information is not understanding
Consider an organisation asking AI how to reduce employee turnover.
The system might recommend better remuneration, improved recognition, flexible work arrangements, and leadership development.
Every recommendation could be reasonable.
But what if turnover is not the real problem?
Perhaps departures are concentrated among employees with critical capabilities. Perhaps one manager is driving resignations. Perhaps low turnover is masking stagnation in a department that urgently needs new skills.
The system may have answered the question correctly without helping the organisation understand the problem.
A better answer to the wrong question does not improve the decision.
It may simply help the organisation move in the wrong direction with greater confidence.
The same distinction applies to prediction.
AI may estimate which employees are likely to resign, which candidates could succeed or which strategic investments might generate the greatest return.
These predictions can be valuable.
But they cannot independently resolve questions of organisational purpose, fairness, acceptable consequences or responsibility.
AI may estimate what is likely. Judgement must consider what matters.
This distinction becomes especially important when decisions affect people, livelihoods, organisational capability or long-term strategic direction.
More intelligence does not guarantee better decisions
It is tempting to assume that combining human and artificial intelligence must produce better results.
Research suggests a more complicated picture.
In a systematic review and meta-analysis of 106 experiments, Michelle Vaccaro, Abdullah Almaatouq and Thomas Malone found that human–AI combinations generally outperformed humans working alone.
However, they did not, on average, outperform the better of the human or AI working independently.
The results also varied by task. Human–AI collaboration showed benefits in content creation but was more likely to underperform the best standalone performer in decision-making tasks.
This challenges an increasingly popular assumption.
Putting a capable person and a capable AI system into the same workflow does not automatically produce superior intelligence.
The outcome depends on how work is allocated, what each contributes and whether the human knows when to trust or challenge the technology.
Another study involving 758 Boston Consulting Group consultants demonstrated what researchers described as AI’s jagged technological frontier.
AI improved performance on tasks within its capabilities. Yet, on a task outside that frontier, consultants using AI were less likely to reach the correct answer than those working without it.
The difficulty is that AI’s boundaries are not always obvious.
A system may perform exceptionally well on one complex task and fail on another that appears straightforward.
Its fluent, confident responses can make these failures difficult to recognise.
The danger is not simply that AI sometimes produces incorrect answers.
It is that people may become increasingly confident in recommendations they are less equipped to evaluate.
Human oversight has little value when the person providing it cannot recognise when intervention is necessary.
When critical thinking changes location
The growing use of AI raises another important question.
What happens to critical thinking when technology increasingly performs the work people previously had to think through themselves?
A 2025 study involving 319 knowledge workers examined 936 examples of generative AI use.
The researchers found that greater confidence in AI was associated with less reported critical-thinking activity, while greater confidence in one’s own abilities was associated with more.
Importantly, the study relied on self-reported experiences. It did not establish that AI causes long-term cognitive decline.
However, it revealed an important shift.
Critical thinking was moving away from gathering information and producing responses towards verifying information, integrating knowledge and supervising AI-supported work.
This changes what organisations need people to do.
Employees must increasingly recognise questionable assumptions, identify missing perspectives, test evidence and determine whether conclusions are appropriate for their circumstances.
Yet these capabilities cannot operate independently of knowledge.
People recognise weak reasoning because they understand the subject. They identify missing evidence because they know what should be present.
An inexperienced employee may ask AI to produce a sophisticated financial, engineering or strategic analysis without having the knowledge required to recognise its weaknesses.
Access to information does not eliminate the need for understanding.
Knowledge becomes part of the infrastructure of judgement.
The organisational risk is that AI reduces the work through which people develop knowledge while simultaneously increasing the importance of applying that knowledge.
Five demands of judgement in AI-enabled work
If organisations want to benefit from increasingly capable AI without surrendering sound decision-making, five human capabilities become particularly important.
Together, they form THINK.
T — Thoughtful Questioning
The first responsibility is to establish whether the organisation is asking the right question.
Thoughtful questioning challenges assumptions, examines how problems are framed and considers which perspectives may have been overlooked.
This goes beyond prompt engineering.
A technically effective prompt can still produce an excellent answer to a poorly defined problem.
H — Human Judgement
Judgement requires weighing evidence against context, uncertainty, competing objectives, and consequences.
It becomes particularly important when several alternatives are plausible or when decisions involve ethical, organisational and human considerations.
AI can inform these decisions.
It cannot remove the organisation’s responsibility for making them.
I — Information Validation
An answer deserves influence only when its evidential foundation can be trusted.
Information validation requires checking sources, testing assumptions, distinguishing facts from estimates, and determining whether findings apply to the situation.
The more convincing AI-generated information becomes, the more important disciplined verification becomes.
N — Navigating Complexity
Organisational decisions rarely affect only one variable.
A productivity improvement may weaken employee development. A restructuring may reduce costs while damaging relationships or institutional knowledge.
Navigating complexity involves recognising interdependencies, competing interests and unintended consequences.
AI can help identify possibilities.
Judgement determines which relationships and trade-offs matter.
K — Knowledge Integration
Information becomes valuable when it is connected with experience, professional knowledge and organisational understanding.
Knowledge integration brings together research, data, practical experience and the perspectives of people affected by a decision.
It prevents organisations from confusing readily available information with sufficient understanding.
Together, these five capabilities represent a different way of thinking about human contribution.
THINK does not position people in competition with AI. It identifies the capabilities required to use AI without surrendering the responsibility to understand, question and decide.
The leadership challenge
Recognising the importance of judgement does not guarantee that organisations will develop or reward it.
Many workplaces continue to emphasise speed, efficiency and visible output.
An employee who challenges an AI-generated recommendation may appear less productive than someone who accepts it immediately.
A manager under pressure may prefer a plausible answer today to a better question that delays implementation.
The organisation may therefore claim to value critical thinking while creating conditions that discourage it.
This creates a difficult leadership paradox.
Organisations could become more dependent on human judgement while doing less to develop it.
The solution is not to preserve human involvement in every task.
Some work can and should be automated.
The challenge is to identify where human judgement remains consequential and deliberately develop the capabilities required to exercise it.
This requires leaders to reconsider several aspects of organisational practice.
1. Distinguish outputs from decisions
An AI-generated report, prediction or recommendation is not necessarily a decision.
Leaders should clarify who has authority to accept, modify or reject consequential recommendations and what evidence is required before acting.
2. Develop knowledge and judgement together
Employees need opportunities to investigate problems, interpret evidence and make decisions with appropriate guidance.
If AI performs all the foundational work, organisations may unintentionally weaken the learning experiences through which future experts and leaders develop.
3. Make questioning and verification part of performance
Employees should be encouraged to challenge assumptions, disclose uncertainty and explain how conclusions were reached.
Performance should reflect the quality of reasoning, not merely the speed or appearance of the final output.
4. Design human oversight around capability
Human involvement should be meaningful rather than ceremonial.
People reviewing AI-supported decisions must have the necessary knowledge, authority and willingness to intervene.
A human signature does not automatically make a decision responsible.
5. Preserve learning and accountability
Organisations should examine whether decisions were appropriately framed, evidence was adequately tested and consequences were considered.
They must also learn from decisions that produce unexpected outcomes.
Accountability cannot disappear into phrases such as “the system recommended” or “the model decided”.
The organisation selected the technology, designed the workflow and acted on the result.
AI can contribute to that process.
It cannot absorb responsibility for the consequences.
Ultimately, the future value of people will not depend on their ability to reproduce information that machines can generate more quickly.
It will depend increasingly on their capacity to understand what matters, challenge what appears convincing and exercise informed judgement when the consequences are significant.
But that capacity must be deliberately developed.
THINK FUTURE Principle
When answers become abundant, value shifts towards the ability to ask what matters, determine what can be trusted and accept responsibility for what happens next.
This principle does not claim that AI will universally increase the economic value of every person or role. It distils a narrower argument: as AI increases access to information, analysis and plausible recommendations, accountable judgement becomes more consequential in deciding what should shape action.
AI may generate the answer in seconds.
It cannot decide whether the organisation has asked the right question, whether the evidence deserves confidence or whether the consequences are acceptable without human objectives, standards and accountability being brought into the process.
The final leadership question is therefore:
If AI can generate an answer in seconds, what is your organisation doing to develop the judgement required to decide whether that answer should shape action?
Coming next
The Boardroom Blind Spot
Boards and executives increasingly examine the financial, operational, cybersecurity and governance implications of AI investments.
Those questions are essential.
One question may still be missing.
What happens to the organisation’s future capability when AI removes the early work through which people once learned, weakens the pathway into expertise or leaves fewer employees prepared for leadership?
In the next essay, we examine an emerging organisational risk:
What if an AI investment improves present productivity while quietly consuming the capability the organisation will need in the future?