– By Jyothi Rani Korem and Manish Gupta
Did you ever come across situations where all student assignments appear strangely similar… too verbose to comprehend, too synthetic to engage, and too generic to resonate with? If so, you are witnessing one of the first visible signs of a profound shift in education. We are living in unprecedented times where artificial intelligence (AI) is increasingly occupying the space once reserved for human intelligence.
Content that takes humans days to learn and process can be generated by AI in minutes, if not seconds. And the pace at which AI capabilities are advancing is only accelerating with each passing day. It has triggered enormous excitement across governments and industries in integrating AI tools into everyday work. It is well evident from McKinsey’s recent survey showing that about 28 percent of companies using artificial intelligence have placed the chief executive officer in charge of AI governance, while 17 percent report that board members oversee it. The same research also reveals that redesigning workflows has the strongest link to realizing profit impact from generative AI.
To match industry expectations, governments are encouraging educational institutions to upgrade themselves. These institutions are doing so by offering fresh electives on AI while also weaving AI into every nook and cranny of their otherwise conventional curricula. While AI is clearly here to stay, educational institutions must pause and remember that for a long time, educational institutions have imparted knowledge in various ways. But thanks to AI, the very relevance of human learning or for that matter even thinking is now under scrutiny, and rightly so. Among the many questions popping into our minds, the one that must concern the policymakers most is rather fundamental: “What should students actually learn in educational institutions?” The answer will shape the future of classroom learning. In short, in an AI-driven world where humans have taken a back seat, the quality of the prompt determines the quality of the result and reiterations become easier than carefully thought-through work. In such a context, education policymakers must exercise restraint and ensure that the rush to adopt AI does not come at the expense of foundational knowledge, a point that we make in this article.
Why the fundamentals still matter
Amid the enthusiasm surrounding generative AI, there is a growing temptation to assume that machines will handle the “thinking” while humans simply supervise the results. If AI can instantly generate answers, why should students spend years mastering mathematics, science, or logic?
The answer is simple: because understanding AI requires precisely those foundations. Education must move beyond rote memorization toward deeper, contextual learning that enables students to interpret algorithmic outputs, evaluate risks, and design responsible human-in-the-loop systems. Workers in an AI-augmented economy will still require strong foundations in mathematics, science, logic, and ethics. These disciplines help individuals recognize flawed reasoning, question automated outputs, and detect when machines might be wrong.
Equally important is systems thinking, the ability to understand how interconnected components interact, how feedback loops amplify or dampen effects, and how complex systems produce unintended consequences. This kind of thinking becomes essential when AI influences decisions in public policy, healthcare, governance, and law. As machines become more sophisticated, the value of human judgment and domain expertise will only increase.
The root cause lies in the way AI works
To understand why foundational knowledge matters so much, it is important to recognize how modern AI systems actually function.
For centuries, technological progress aimed at precision. Mechanical clocks, factory machines, early calculators, and spreadsheets all operated on deterministic principles: the same input produced the same output every time. Generative AI represents a fundamentally different technological paradigm. Most contemporary AI systems operate through probabilistic inference. Instead of calculating a single correct answer, they predict likely outcomes based on patterns in past data. Their responses are therefore not objective truths but statistical predictions generated by models trained on enormous datasets. This distinction has profound implications for decision-makers.
To understand the limitations of assistive AI, we use the acronym ABHAV, representing five recurring challenges: Assumption, Blindness, Hallucination, Ambiguity, and Validity. First, assumption. AI-generated responses often appear persuasive but may rely on superficial patterns rather than grounded reasoning. Because large language models operate through next-word prediction, the outputs can sound authoritative while lacking real-world applicability. Second, blindness, or context oversight. AI systems learn from existing data and often struggle with cultural nuance or local context. Idioms, geographical differences, or historical references may be misinterpreted. One recent study suggests that around 14 percent of generative AI outputs lack sufficient contextual validity. Third, hallucination, where AI systems confidently generate fabricated information. There have already been examples of legal briefs citing court judgments that never existed or research reports referencing imaginary sources. Fourth, ambiguity. For complex questions with multiple possible interpretations, such as strategic or policy decisions, AI systems may generate different answers each time the same question is asked. Finally, validity reflects the probabilistic variability of AI outputs. The same prompt can produce starkly different summaries or perspectives, creating inconsistencies that may confuse decision-makers. These limitations do not mean AI is ineffective. But they highlight why human understanding remains essential.
What should education policymakers do?
For policymakers, particularly those shaping national education systems, the challenge is not whether AI should be adopted. AI is already becoming an indispensable part of modern economies. The real challenge is preparing students to work intelligently with AI rather than simply rely on it. One priority should be teaching students to treat AI outputs as hypotheses rather than final answers. AI-generated content should be accompanied by references, evidence, and verification processes.
Second, education systems should emphasize contextual knowledge. AI systems require detailed information about geography, culture, and historical context to produce meaningful insights. Without such understanding, AI risks offering generic solutions to complex local problems. Third, organizations and governments may need new oversight mechanisms such as cross-disciplinary review teams to evaluate AI-assisted decisions. Despite its limitations, assistive AI is already producing valuable outcomes. In healthcare, institutions such as Mayo Clinic use AI in radiology to analyse imaging data alongside patient histories, enabling faster diagnosis and improved clinical decisions. In manufacturing, companies like Siemens deploy AI to predict equipment failures and optimize maintenance.
India itself is entering the era of agentic AI, where systems can execute complex tasks with elements of reasoning and memory. More than 80 percent of Indian organizations are currently exploring autonomous AI agents. Yet the talent pipeline remains a concern. Forecasts from Bain & Company suggest India may face a shortage of more than one million AI-skilled workers by 2027, even as Deloitte estimates that the country’s AI talent pool could reach 1.25 million. This gap underscores a crucial truth: technological progress cannot succeed without strong educational foundations.
The way forward
Ultimately, the question facing society is not whether AI will replace human intelligence. The real challenge is learning how human expertise and machine intelligence can complement each other. AI should be treated as assistive, not authoritative. Its outputs should serve as inputs for human deliberation rather than final decisions. This requires decision processes that actively check the limitations captured in the ABHAV framework by involving domain experts, stakeholders, credible evidence, and professional standards.
AI-assisted decision-making remains in its infancy. While probabilistic models can dramatically improve efficiency, effectiveness still requires human supervision, interpretation, and judgment. In a world increasingly shaped by intelligent machines, the most valuable skill may not be the ability to generate answers quickly. It may be the ability to understand when those answers are wrong. And that is precisely why strengthening the fundamentals of education has never mattered more.
– The author are Professor of Practice, School of Management, Mahindra University and Associate Professor, School of Management, Mahindra University
DISCLAIMER: The views expressed are solely of the author and ETEducation does not necessarily subscribe to it. ETEducation will not be responsible for any damage caused to any person or organisation directly or indirectly.
