I grew up in a family where education was not a phase of life. My father navigated the world's oceans and learned constantly — every port, every route, every unexpected weather system demanded new knowledge, new adaptation. My mother taught in a classroom and learned alongside her students as the world they were preparing them for kept changing. My grandmother ran a school and understood, intuitively, that the job of an educational institution is not to transmit a fixed body of knowledge but to build the capacity to keep learning after the institution's walls are left behind.
The model of learning as a defined period — something that happens in childhood and young adulthood, in designated institutions, and then concludes when work begins — was always a simplification. It captured something real about where formal education happens. But it missed something important: that the most consequential learning in a person's life often happens after formal education ends, in workplaces, in communities, in the confrontation with problems that no curriculum anticipated.
Artificial intelligence is now ending even the simplification. The skills that confer economic security today are not the skills that will confer economic security in ten years. The World Economic Forum estimates that 50% of all employees will need significant reskilling by 2025 — a figure that has been updated upward with each successive iteration of its Future of Jobs Report.1 The IMF projects 92 million job displacements by 2030, alongside 170 million new roles — a net positive, but only for those who can navigate the transition.2 And navigation requires continuous learning.
"The model of learning as something that happens in a defined period of life before work begins is breaking down. AI makes lifelong learning not just possible but necessary."
The argument for AI as an enabler of continuous learning is real and important. AI tutoring systems can now adapt to individual learners in ways that formal educational systems — constrained by the economics of class sizes, curriculum standardisation, and fixed timetables — structurally cannot. They can identify gaps in understanding with a precision that a teacher managing thirty students cannot. They can deliver learning at the moment when it is needed — not at the moment when a course is scheduled. They are available at three in the morning for the nurse who works night shifts and is studying for her next qualification, for the factory worker who is learning new technical skills in the time between the end of his shift and the beginning of his family's evening, for the student in rural Karnataka who has no local institution offering what she needs to learn.
My doctoral research found that AI learning systems, when designed with trust and transparency as primary values, produce measurable improvements in both adoption and learning outcomes across enterprise contexts.3 The technology works. The question is not whether AI can enable continuous learning. It is whether the systems, institutions, and policies surrounding AI can support the human reality of learning continuously — with all the complexity, irregularity, and emotional texture that continuous learning involves.
The possibility that AI creates for individual learners is running well ahead of the institutional capacity to support it. The gap manifests in at least four distinct dimensions.
Learning that happens outside formal institutions — in AI-enabled programmes, in workplace training, in community contexts — is still not consistently recognised by the credential systems that employment and further study require. The learning may be real; its recognition is not guaranteed.
Most public funding for education is structured around initial formation — the years of formal schooling and higher education. There is no equivalent public investment structure for the continuous reskilling that the AI transition requires of working adults throughout their careers.
Continuous learning requires time. Most working adults, especially those in lower-income employment, do not have discretionary time that can be allocated to learning. The technology that makes learning more accessible does not, by itself, create the time that learning requires.
AI-enabled learning requires connectivity, devices, and a sufficient baseline of digital literacy to engage with the tools. These are not uniformly distributed — within countries or between them. The populations who most need continuous learning support are often the least equipped to access it.
Addressing these gaps requires not just better technology or better content but a different institutional imagination — one that takes seriously the question of what it would mean to build an economy and a society genuinely oriented around continuous learning.
Some of what this requires is already being discussed in policy circles. Portable learning accounts — individual learning credits that follow workers across employers and career transitions — have been piloted in Singapore, France, and the United Kingdom with varying degrees of success.4 Recognition of prior learning — formal mechanisms for credentialing knowledge and skills acquired outside formal institutions — is embedded in several national qualification frameworks but remains inconsistently applied. Employer obligations for worker reskilling — requirements that organisations deploying AI provide the training required for workers to adapt — are beginning to appear in some regulatory frameworks.
These are partial answers to a large question. The fuller answer would require a much more fundamental reimagining of the relationship between learning and working — one that treats learning not as preparation for work but as an ongoing dimension of work, with the institutional and economic infrastructure to support it.
I want to say something about the human experience of continuous learning that policy discussions often miss. Learning is not only a technical process of skill acquisition. It is an experience that can be affirming, disorienting, humbling, joyful, and exhausting — often all at once. For a person who left formal education decades ago, returning to a learning context involves not just the cognitive challenge of acquiring new knowledge but the emotional challenge of being a beginner again, in a world that may not have made that experience welcome the first time.
AI learning systems that are designed only for efficiency — that maximise knowledge transfer without attending to the emotional and relational dimensions of learning — will not serve the full range of people who need them. The evidence from human-centred design research is clear: learners who feel seen, understood, and supported in their learning experience persist longer, engage more deeply, and achieve better outcomes than those who encounter only technically competent but emotionally inert systems.5
This is why, in my research and in my teaching, I keep returning to the conviction that the most important design question in AI-enabled learning is not "how smart is the system?" but "how well does the system understand the person?" The answer to that question — embedded in the system's architecture, in its feedback mechanisms, in its treatment of the learner's context and history — determines whether continuous learning is a possibility that the system enables or a burden it adds to lives already stretched thin.
"The question is whether our systems — educational, organisational, political — are ready to support the human reality of learning continuously, with all the complexity and emotional texture that continuous learning involves."
I am writing this essay as someone who is herself a continuous learner — who spent years in industry before returning to doctoral research, who teaches while researching, who is learning from students and participants and policy conversations even as she contributes to them. The experience of continuous learning is not abstract for me. It is the texture of my working life.
What I know from that experience is that the conditions that make continuous learning possible are not primarily technological. They are relational, institutional, and cultural. They require someone who believes you can learn — a teacher, a mentor, an institution, a culture. They require time and space that is structured around learning rather than structured around everything else. They require a tolerance for the discomfort of not knowing, which is the unavoidable starting point of every genuine learning experience.
AI can help create some of these conditions. It cannot create all of them. The challenge of making learning genuinely continuous — for everyone, across a lifetime — is not primarily an AI design challenge. It is a human and institutional challenge of the first order. And it is one we have not yet met.
Are we ready? Not yet. But the gap between what is possible and what is in place is not fixed. It is a design problem. And design problems have solutions.
The six ideas explored in this series — trust, literacy, governance, accessibility, collaboration, and continuous learning — are not separate topics. They are the same question, approached from different angles: how do we ensure that artificial intelligence expands human potential without leaving anyone behind? I keep returning to each of them because I do not think the question has a final answer. But I believe that asking it carefully, persistently, and with genuine attention to the people for whom the stakes are highest, is how we build the future this moment requires.