Why leaders need to question reality in the age of AI

In a recent Harvard Business Review (HBR) article, the authors make a case that is both correct and overdue: Leaders now make philosophical decisions, not just financial ones.

Adopt an AI model, and you inherit premises about truth, agency, and value that someone else wrote into the tool. Anthropic’s published constitution for its model, Claude, is only the most transparent instance. As the social consensus that once settled questions of speech, expertise, and obligation fractures, leaders can no longer outsource them to a stable culture. They must reason in the open.

So, their recommendation to build proficiency in ontology, epistemology, and ethics is sound. I want to push it one layer further, using a tradition that has audited exactly these three questions for over two thousand years: Advaita Vedānta, in the precise lineage in which I work. It accepts the three domains and adds what the modern frame omits: Before asking what is true or right, it asks how real the thing in front of you is. Most leaders, most of the time, react with full conviction to things that are not there in the way they take them to be.

Ontology: Not “what is it?” but “how real is it?”

The HBR authors use Apple’s stance on customer data: Is it raw material to be extracted or the embodiment of an entrusted relationship? Choosing between those definitions, they say, is the philosophical work.

It is the easy half. Both are legitimate, stable, public construals. In Vedāntic terms, both are vyāvahārika: Real at the level of transaction, holding up across observers and time. The error that wrecks judgement lives one level down, in the prātibhāsika, the merely apparent, real only for me, only now. The classic image is a rope on a dim path mistaken for a snake: The fear was real; the snake never was.

A startling amount of what drives leaders has this status. The founder whose worth rises and falls with the valuation has fused their identity to a number that is vyāvahārika as a financing term and prātibhāsika as a measure of the self, and at the down-round discovers that they were fighting a snake. The certainty that a regulator acts in bad faith, that a competitor’s move was aimed at them, these read as features of the world and are often projections laid over it. The term is adhyāsa: Superimposition, casting what is not there onto what is, then responding to the composite as if given.

The discipline is not rising above the relative world but grading correctly within it, asking of any certainty driving a decision: Is this vyāvahārika, or is it prātibhāsika, something I am projecting and have mistaken for real?

The HBR question assumes the contents of the leader’s mind are all candidates for deliberate selection. The prior question is which are even there.

Epistemology: Knowing the reach of each instrument

Here the two frames nearly converge. Vedānta inherits a careful inventory of pramāṇas, valid means of knowing, each authoritative only in its domain. Perception (pratyakṣa) knows the perceptible and is worthless beyond it; inference (anumāna) reaches further but rests on what perception supplied; testimony (śabda) conveys what you cannot perceive or infer and is only as good as its source. The discipline is to know, for any claim, which instrument it rests on and whether that instrument can bear the weight.

Toyota’s genchi genbutsu, “go and see for yourself”, is not folk wisdom but rigorous epistemology: A refusal to let a report stand in for the perception it was substituting. It is a ruling about pramāṇa.

The same lens exposes the category error in the Pentagon’s 2026 demand for AI that is “objectively truthful” and free of ideological contamination. A model’s output is a form of testimony, śabda, which carries the standpoint of its source by nature; you cannot purify it into perception through procurement language. The exposed leaders are those who never ask on what means of knowing a claim rests, granting model output the certainty of perception or, once burnt, distrusting every instrument equally.

Calibrated confidence, full trust in each instrument within its range, none beyond, is the whole of epistemic proficiency, and it is teachable.

Ethics: Reading the field, not consulting the rulebook

The HBR test is memorable: Would you hold a supposedly ethical position at a cost of 10 percent of revenue? If not, it was branding.

It is a fine test of sincerity and no test of correctness.

A leader who would burn 10 percent defending a position can be sincerely, expensively wrong. The test is one a fanatic passes most easily. In the modern frame, an ethical commitment is constructed and defended: Pick your values, hold the line. In the Indian frame, the ethical order, dharma, is not constructed at all. It is closer to something read: A discernment of what this situation, with these relationships and stakes, calls for from a person in my specific position. And it is not free-floating.

Dharma is one expression of a larger lawful order, the same intelligence that makes the world a cosmos rather than chaos, which the tradition calls Īśvara: Not a deity issuing decrees but the total intelligent order itself, of which the moral order is one strand. That is what gives dharma its objectivity. It is read off the order, not asserted by the will, which is why two sincere leaders can both misread it.

That qualification, “in my specific position”, carries weight. Dharma is not a universal maxim but svadharma, the obligation arising from one’s actual role. A CEO’s, a board member’s, and a regulator’s obligations diverge, even when they face identical facts. The question is not “what is my conviction and will I pay for it?” but “have I correctly read what this situation requires of someone standing where I stand?”

This is why the Bhagavad Gītā opens not with a doctrine but with a leader in crisis. Arjuna, a commander, collapses at the decisive moment, and the answer is not a rule but an analysis of his svadharma, of the intelligent order his action sits within, and of the disposition from which he must act: Fully committed to what the situation demands, while releasing the grip on an outcome he cannot control.

That disposition, karma-yoga, is not withdrawal or indifference. It is acting wholeheartedly and holding the result lightly, the actual source of steadiness under fire. A commitment defended out of attachment to outcome is the brittle kind that flips when the outcome turns; the consumer-brand value reversals of recent years are the textbook case. Conviction held out of rāga-dveṣa, the pull of like and the push of dislike, is not a still centre. It is a weathervane mistaking itself for a compass.

What the separation conceals

I walked the three rooms in the modern frame’s order. But that ordering assumes the three are separate enquiries, taken one at a time. The Indian tradition does not carve the territory this way.

The three are faces of one question: What is the case (ontology), how is it known (epistemology), and what does it ask of me (ethics) are inseparable because one act of clear seeing answers all three, and one self-deception corrupts all three. A leader who misgrades a prātibhāsika projection as real will reason about it with false certainty and act on a misread of what is demanded: Not three failures but one. The gain is equally singular: A leader operating from a wider, correctly graded picture is, in one motion, less reactive, more accurate, and steadier.

This requires no exotic commitment. It all happens within the transactional order in which business is conducted; it is available to a thoroughgoing pragmatist, a finer set of instruments for the world as it is.

There is, in this tradition, a further inquiry into the nature of the one doing the seeing, knowing, and acting; the nature of the world that is seen; and the single cause from which both knower and known arise. The modern frame leaves all three unexamined, looking outward at the business and never asking after the seer, the seen, or their ground. That road is longer, and not the subject here. None of the gains above depend on travelling it. The deeper questions are real, and they can wait. The wider picture is enough to lead from.

In fairness, the modern frame has real advantages this register does not match: Analysing institutional power, making an argument publicly defensible to those who do not share your premises, and providing the formal machinery of valid reasoning. A complete leader wants both. But on how real the things driving your decisions are, the older tradition is holding an instrument the newer one has not yet built.