Adarsh Sudhindra is the Chief Innovation Officer at Excelsoft Technologies, where he leads AI and product innovation.
getty
For two centuries, the classroom has been an answer factory. Industrial-era schooling was engineered for a world in which facts were scarce, books were expensive and finding something out could take a day in a library. So, we optimized education for storage and retrieval: memorize, recall, reproduce. It was a rational design for its time—the economy hired the person who knew things.
I have spent my career building learning and assessment technology, and I believe the scarcity that system was built on has now inverted. AI produces competent answers on demand—summaries, explanations, procedures, working code—at near-zero cost. What remains scarce is everything the answer factory never prioritized: framing the right problem, weighing conflicting evidence, making decisions under uncertainty and taking responsibility for outcomes. In my view, an education system that keeps certifying humans for skills machines now give away is not preserving rigor; it is manufacturing obsolescence.
The Premium Moves Up The Stack
Educators have long invoked Bloom's taxonomy, the familiar hierarchy that climbs from remembering facts toward analyzing, evaluating and creating. But the timetable told the truth: Most classroom hours were at the bottom because remembering was the cheapest to teach and the easiest to test at scale. AI rewrites those economics. The bottom of the hierarchy is now effectively automated, and the premium moves to the layers we long treated as enrichment.
There is a subtler shift underneath. Knowledge work is changing from producing answers to choosing among them. An AI will offer five fluent options; the human's job is to notice which one is wrong for reasons the model cannot see—context, ethics, second-order consequences. The hazard of this era is not ignorance. It is confident, fluently formatted error. That is why I see critical thinking as a workplace necessity rather than an academic virtue and decision-making as the most employable capacity education can build.
Automate The Clerical, Never The Cognitive
"Let AI do the mundane" is the right instinct, but it hides a trap because two very different things in education look mundane. The first is clerical drudgery: grading mechanics, paperwork, scheduling and progress reports. Working with schools, universities and assessment bodies, I have watched this load consume evenings teachers would rather spend teaching—and it teaches no one anything. Automate all of it, and return those hours to the work only humans do well: mentoring, questioning, noticing the student who has quietly stopped trying.
The second is difficulty: the friction of wrestling with a problem you cannot yet solve. That is not drudgery; that is the mechanism. Cognitive effort is to thinking what resistance is to muscle. If AI removes the struggle during the years judgment is being formed, we will graduate students who have outsourced the very capacity education exists to build. The design principle I hold to is easy to state and demanding to live by: Automate the administration of learning, never the labor of thinking. AI should carry the schoolbag, not climb the hill.
Assessment Is The Steering Wheel
Nothing upstream changes until the test does. Schools teach what examinations reward, and students practice what rubrics count. As long as we assess recall and standard procedure, classrooms will keep teaching machine skills to humans—and AI will keep "helping" in ways we label cheating. The deeper problem is not that students delegate the task; it is that the task was worth delegating in the first place.
The most promising redesigns I have seen reward what cannot be handed off to a machine: open problems with more than one defensible answer, oral defense of reasoning, sustained projects and critique exercises that grade students on identifying where an AI-generated answer fails. When the test measures judgment, classrooms grow judgment.
What Leaders Can Do Now
Three moves strike me as most urgent. First, audit the automatable hours in your institution or learning function and reinvest them in discussion, projects and inquiry rather than more content coverage. Second, equip teachers before students; classrooms change when educators become confident orchestrators of these tools rather than reluctant police. Third, teach AI literacy as interrogation, not operation—the skill is to treat a model's output as a claim to verify, probing provenance, bias and missing context, rather than as an oracle to transcribe.
Education's Original Job
The fact-based school was never the ideal; it was a compromise with scarcity. Long before facts were cheap, education meant dialogue, argument and judgment. Socrates never handed out a worksheet. AI does not threaten that older tradition; it ends the compromise that interrupted it. We taught generations of humans to answer like machines because we had no machines. Now we do. The takeaway for anyone who leads learning—in a school system or a company—is to move the human hours up the stack: Hand clerical work to machines, protect the productive struggle and rebuild assessment until it rewards thinking. AI has made answers cheap. The work of education now is to make thinking valuable again.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?