Will AI Replace Jobs? Preparing Children for a Radically Different Future of Work

Originally published in May 2025. Substantially updated in September 2026 to reflect advancing AI capabilities, emerging labour-market evidence and the growing importance of protecting children’s independent thinking and learning.

When SAIFCA first published this article in 2025, we warned that today’s children were likely to enter a dramatically different workforce and that education, public policy, and wider assumptions about work needed to change in response. 

Since then, AI capabilities have advanced substantially, use across workplaces has expanded, and early signs of labour-market disruption have become more visible. We are therefore presenting this argument with a stronger framing. 

Although the pace and eventual scale remain uncertain, we now consider severe disruption to employment and established routes into work as an increasingly plausible outcome for which governments, education systems and families should be actively preparing.

While this doesn’t mean that every job will disappear, that widespread permanent unemployment is inevitable, or that there is no possibility of a positive future, it does mean that relying on reassuring historical comparisons or waiting for conclusive evidence could leave children facing changes for which our institutions have, so far, not done enough to prepare them.

The choices made now will help to determine whether AI supports a fairer and more secure future, or concentrates wealth and opportunity while weakening the pathways through which young people can build skills, independence, and thriving adult lives.

Key Points

• AI is likely to transform a significant proportion of existing work, although exposure to AI does not automatically mean that an entire job will disappear.
• The exact timeline is uncertain because technical capability, cost, infrastructure, reliability, organisational adoption, regulation, and public acceptance do not advance at the same rate.
• Significant disruption can occur without every occupation being automated. If AI allows fewer workers to produce the same output, employment opportunities can decline even while people remain involved.
• Entry-level jobs could be particularly vulnerable because they often contain routine cognitive tasks that AI systems can increasingly perform. These roles have traditionally allowed young people to gain the experience needed for more advanced work.
• New jobs will be created, but we cannot assume that they will emerge quickly enough, exist in sufficient numbers, offer equivalent security, or be accessible to those whose opportunities are displaced.
• Children need strong knowledge, metacognition, independent reasoning, creativity, communication, collaboration, and practical experience. Simply teaching them to use AI tools will not provide adequate preparation.
• Excessive cognitive offloading to chatbots could weaken the very capabilities that become most important in an AI-shaped workforce.
• Parents and schools can take practical action, but responsibility cannot be placed on them or on children to adapt to disruption created by decisions made elsewhere. Governments, employers, and AI companies must also act.

A Changing Workforce Is No Longer a Distant Possibility

Predictions about AI and employment can vary widely. Some foresee mass unemployment, while others expect AI to increase productivity and create more jobs than it removes. Neither outcome should currently be treated as certain.

There are, however, strong reasons to expect major disruption.

Disruption to employment does not depend on AI systems being able to perform every part of a profession perfectly, insofar as most jobs consist of many different tasks - and if a system can undertake enough of those tasks quickly and cheaply, an organisation could reduce recruitment, combine roles, outsource less work or expect a smaller number of employees to produce considerably more.

This is already placing pressure on some areas including administration, customer service, translation, software development, design, research, marketing, and some forms of analytic work. Work requiring physical presence, complex human relationships, accountability, or extensive knowledge of a particular setting, could potentially be much more difficult to automate, although developments in robotics and AI agents could feasibly change this.

The International Labour Organisation’s 2025 assessment of occupational exposure found that around one in four workers globally were in occupations with some exposure to generative AI. The ILO concluded that transformation was more likely than complete replacement under current conditions, while also recognising that exposure had increased as AI capabilities developed.

The World Economic Forum’s Future of Jobs Report 2025 estimated that the major forces reshaping the global economy could create or displace the equivalent of 22 per cent of current formal jobs by 2030. Its employer survey anticipated both job creation and job displacement from AI, alongside extensive changes to the division of work between people and technology.

These figures are projections, and they also tend to assess identifiable uses of existing or near-term technology, rather than every consequence of further capability advances, which is largely unknown. A large part of their value lies in demonstrating that significant labour-market change is no longer confined to the most dramatic forecasts.

Why Timelines Remain Uncertain

AI capabilities can advance rapidly, while workplaces and public institutions usually change much more slowly.

Employers need time to reorganise processes, test systems, train staff, manage legal risks, and integrate new technology with existing infrastructure. AI systems can also remain unreliable in ways that limit their use in higher-stakes settings. Regulation, public resistance, professional standards, and the value placed on human contact could all slow or shape deployment.

Economic effects can take longer to become visible than technical capabilities. Businesses do not always respond by announcing that employees have been replaced by AI - change can occur through recruitment freezes, smaller graduate intakes, reduced freelance work, roles left unfilled after someone departs, or rising expectations of what each remaining employee should produce.

This creates a dangerous possibility - that society could mistake a delay between capability and widespread adoption for evidence that serious disruption will not happen. 

The Risk to Young People

The ILO’s Global Employment Trends for Youth 2026 describes an increasingly difficult transition from education into secure work. It estimates that 6.1 per cent of jobs currently held by people aged 15 to 29 fall within the categories most exposed to AI. If only 10 per cent of those exposed jobs disappeared completely, 5.6 million young workers could face unemployment, a change of occupation or departure from the labour market.

AI is not the sole cause of current difficulties facing younger workers - slower economic growth, demographic changes, geopolitical instability, and longstanding weaknesses in youth employment all contribute. Even so, AI could intensify these pressures, especially if it reduces the availability of junior work.

Entry-level roles often include research, drafting, basic analysis, administration, customer contact, and other tasks through which less experienced workers learn. If organisations automate those tasks or give them to a smaller number of experienced employees using AI, young people could encounter a missing first rung on the career ladder.

This raises a significant problem, and we cannot solve it by telling children to become more highly skilled - that experienced professionals possess judgment, organisational knowledge, and practical understanding, partly because they once undertook junior work. If fewer people gain that experience, the future workforce could eventually face shortages of genuinely capable senior workers, while immediate demand for junior staff also declines.

The effects will also be unequal. Children with access to strong education, professional networks, financial support, and opportunities for practical experience will be better placed to navigate repeated transitions. Those already facing disadvantage could be directed towards narrowing pathways, or offered training for roles that disappear before they have time to establish themselves.

Preparing Children Requires More Than Teaching Them to Use AI

AI literacy will be important. Children should understand how AI systems work at an appropriate level, where they can be useful, how their outputs can mislead, and how their use affects privacy, fairness, safety and other people.

Yet AI literacy risks being interpreted as proficiency in using chatbots and other AI tools, which is far too narrow.

Employers will undoubtedly expect many workers to use AI. But a child who has been encouraged to depend on AI for thinking, writing, planning, and problem-solving, will not be well prepared.

Effective use depends on the human user having enough knowledge and judgment to direct the task, notice weaknesses, challenge an answer, and recognise when the system should not be used.

Children therefore need a combination of strong foundational and subject knowledge, with wider capabilities such as reasoning, creativity, communication, collaboration, self-awareness, and adaptability. They also need metacognition.

Why Metacognition Will Become More Important

Metacognition involves understanding and regulating our own thinking. It includes being able to:

• Decide how to approach an unfamiliar problem.
• Recognise what we know and what we do not know.
• Monitor whether a strategy is working.
• Notice confusion, gaps, and contradictions.
• Check evidence and revise an initial conclusion.
• Reflect on what helped us learn and what should be done differently next time.

These abilities become especially important when working with systems capable of producing fluent and convincing answers within seconds. Fluency can create an impression of understanding or reliability that the content does not deserve. A person needs sufficient knowledge and metacognitive awareness to recognise when an answer is incomplete, misleading, or outside their ability to verify.

Metacognition also supports adaptation. We cannot know which specific technical skills will remain valuable throughout a child’s working life. Someone who understands how they learn, can assess their own limitations, and can build new knowledge deliberately, will be better equipped to respond as roles change.

However, this should not be mistaken for an argument that subject knowledge has suddenly become less important - critical thinking cannot be reliably applied to information a person knows almost nothing about. General capabilities need to be developed through knowledge-rich learning, repeated practice, and experience across different contexts.

Cognitive Offloading and the Developmental Risk of Chatbots

People have always used tools to reduce mental effort. For example, calculators, search engines, and navigation systems all allow forms of cognitive offloading, and this can be helpful when a tool removes unnecessary burden or frees a person to concentrate on a more demanding part of a task.

Generative AI can go much further by producing the plan, reasoning, explanation, wording, and final answer. Used without sufficient care, it can remove the mental activity through which the child was supposed to learn.

The most important questions concern which part of the thinking process has been handed over, whether the child has already fully developed the relevant capability, and whether they can understand and evaluate what the system produces.

A child who first attempts a problem, explains their reasoning, then uses AI to test or extend it, is having a different experience from a child who asks a chatbot for the answer and quickly edits the resulting text. The finished work could look similar while representing very different levels of learning. And we don’t currently know how either is shaping long-term cognitive development.

Emerging research needs careful interpretation, but it already provides grounds for concern. A 2025 study of knowledge workers found that greater confidence in generative AI was associated with lower reported critical-thinking effort. This study concerned adults, relied partly on self-reporting and did not establish long-term cognitive decline. Research involving children and sustained use is still developing.

But we should not wait for evidence of measurable long-term harm before protecting the learning processes known to build knowledge, reasoning and independence. 

Children need to experience uncertainty, effort, mistakes, revision, and the satisfaction of reaching understanding. If an AI system routinely intervenes before those processes occur, apparent productivity can disguise lost development.

What Parents and Carers Can Do

Parents cannot predict the exact careers that will remain secure, and they should not feel responsible for individually solving a potential global economic transition. But they can still help children develop a strong and resilient foundation. These ideas can be adapted and applied at an age-appropriate level: 

What Schools and Educators Can Do

Schools have an increasingly important role in protecting the development of capabilities that could be obscured by polished AI-generated work, and helping to prevent over-reliance on AI.

What Governments and Policy Leaders Must Do

The scale of the possible transition requires structural action - preparing children for disruption cannot be reduced to personal resilience, career choices, or the responsibility to keep retraining. 

We Need to Shape the Transition

AI could support scientific discovery, improve services, remove dangerous or repetitive work, and contribute to wider prosperity. But these potential benefits do not guarantee a fair labour market, or remove the possibility of serious disruption.

History shows that technological change can create new forms of work - it also shows that transitions might be prolonged, unequal, and harmful to particular communities. AI differs from many previous technologies in its ability to perform a widening range of cognitive tasks across multiple sectors, and in the speed at which its capabilities can advance.

We should not promise children that familiar careers will remain available, nor tell them that individual adaptability will always be enough. They need strong minds, wide capabilities, human relationships, and meaningful opportunities to develop competence and confidence. They also need adults and institutions willing to make difficult decisions about how AI enters education, employment, and society.

The future of work is not determined by technical capability alone. It will also be shaped by regulation, employer choices, educational priorities, and decisions about how the benefits and burdens of automation are shared.

SAIFCA’s Full Guide to Understanding AI Risks to Children

Introducing our new guide for parents and educators - explaining all the key risks to children from AI, and what you can do

The Safe AI For Children AllianceSAIFCA