I want to start with a number that should unsettle every leader in this room: 44%.
That is the share of professionals who begin doubting their own judgment the moment an AI system disagrees with them. Not 4%. Not 14%. Forty-four percent. Nearly half your strategy team, your finance function, your product leaders — the people you are paying for their independent thinking — will second-guess themselves the instant a model pushes back.
Now layer this on top of a Harvard Business Review study published last week, co-authored by Leonid Sudakov — 25 years in senior leadership at Mars, Danone, and PepsiCo — and INSEAD strategy professor Nathan Furr. In a rigorous field experiment, 228 experienced evaluators assessed 48 submissions to an MIT global innovation challenge. Some evaluated alone. Others had access to AI outputs. The result was not what the AI optimists expected.
The group with AI wasn't more accurate. They were more uniform. What the AI did — in one of the most carefully designed business school experiments of 2026 — was compress the range of original human thinking. The evaluators who used AI reached more similar conclusions, covered less perceptual ground, and formed fewer independent views. They had more intelligence. They exercised less judgment.
If that can happen in a controlled experiment with motivated, experienced professionals reviewing 48 submissions, ask yourself what is happening in your own strategy team's Monday morning meetings — where the AI output is on the screen before the conversation starts, and the social pressure to agree with the machine is invisible but real.
THE JUDGMENT GAP — WHAT THE RESEARCH TELLS US
FINDING
STATISTIC
Professionals who doubt their own judgment when AI disagrees
44%
Who turn to AI first under uncertainty
61%
Who rely on AI under time pressure
58%
Who use AI to validate their own reasoning (not explore it)
51%
Who say they feel MORE independent without AI assistance
70%
Evaluators in the Harvard/MIT experiment
228
MIT innovation submissions assessed
48
THE TWO CAPACITIES THAT ARE QUIETLY DISAPPEARING
Sudakov and Furr name two specific cognitive abilities that AI is eroding in leaders — and once you know what to look for, you will start seeing the erosion everywhere.
The first is breadth of perception: the ability to notice weak signals, adjacent patterns, and information that sits just outside the obvious frame. Great strategists, great investors, and great executives have always differed from competent ones by what they notice first — the anomaly in a market report that everyone else skims past, the customer complaint that reveals a product category nobody has built yet, the geopolitical shift that translates into an acquisition opportunity eighteen months before the rest of the industry sees it. AI systems, by design, surface what is statistically probable given the training data. They are excellent at the centre of a distribution. They are blind to its edges.
The second is independence of interpretation: the ability to form a view that differs from the prevailing model output — and to hold that view under social and institutional pressure. This is the hardest capacity to maintain, because AI disagreement has the appearance of objectivity. When the model says X and you believe Y, the burden of proof feels like it sits with you. Over time, leaders who face this choice repeatedly start choosing X. Not because they are wrong. Because the cost of being wrong and different is higher than the cost of being wrong and aligned with the machine.
"As organizations gain more intelligence with AI tools, leaders are being trained out of the very capacity that creates competitive advantage — original judgment."
— Leonid Sudakov & Nathan Furr, Harvard Business Review, August 2026
The Human Clarity Institute's 2026 Decision-Making Report captures the systemic version of this problem in a phrase that I haven't been able to stop thinking about: the "Perceived Control Gap." Executives believe they are in control of AI-assisted decisions — 74% say they feel in control. But 70% of the same cohort report feeling more independent when they make decisions without AI assistance. Control and independence are diverging. Executives are retaining responsibility for outcomes while ceding the cognitive territory that makes the outcomes genuinely theirs.
WHY THIS IS A COMPETITIVE STRATEGY PROBLEM, NOT AN HR PROBLEM
Here is the argument I want to make directly, because I've seen it play out with real businesses in my own portfolio this year.
When every executive team in your industry uses the same AI tools, trained on the same data, the outputs will converge. Your competitive intelligence function and your competitor's will reach the same conclusions about market trends. Your pricing model and theirs will hit the same optimal zones. Your M&A screening and theirs will identify the same acquisition targets. What looks like better decision-making — faster, more data-rich, more defensible — is actually the systematic elimination of the differentiation that makes one C-suite more valuable than another.
The executives who will create asymmetric value in the next five years are not those who use AI best. They are those who retain the capacity to notice what AI misses and hold views that AI cannot generate. That is now a competitive moat. And like most moats, it requires active investment to maintain — because the default trajectory, without deliberate effort, is erosion.
THE BOARDROOM RISK
If your board's strategic discussions begin with an AI-generated briefing, you have built a structural bias toward the model's priors into every major decision. The board sees what the model found important. What the model didn't find — by definition — never enters the room. This is not a technology risk. It is a governance risk.
WHAT INDIA AND UAE LEADERS SHOULD PAY PARTICULAR ATTENTION TO
There is a dimension of this research that particularly matters to the CXO Magazine audience — executives operating across India, UAE, and the Gulf — and it is one that the Western-centric HBR framing glosses over.
AI systems are trained predominantly on Western corporate data. Their priors about market structure, competitive dynamics, consumer behaviour, and strategic opportunity reflect that dataset — not the realities of a Tier-2 Indian city, an emerging NRI investment corridor, or a Gulf state undergoing Vision-driven economic transformation. When an Indian executive defers to an AI model's market assessment, they are frequently deferring to a system that has never encountered the market they are operating in.
The independent judgment gap is wider here, not narrower. The executives who will dominate in India's PropTech, FinTech, and AI education sectors over the next decade are those who synthesise AI-generated intelligence with local pattern recognition — knowledge the model cannot have. That synthesis requires the two capacities Sudakov and Furr identify: broad perception and independent interpretation. Both require active cultivation.
WHAT TO DO BEFORE YOUR NEXT STRATEGIC CYCLE
The research proposes two organisational interventions. I want to give you the practical version of each — the one that works in a real executive team, not a Harvard seminar.
01. Institute pre-AI judgment sessions before any strategic decision
Before the AI briefing goes to the leadership team, require every key stakeholder to write — in two to three paragraphs — their independent read of the situation. Not what they think the answer is. What they notice. What feels off. What the data they already have suggests before the model weighs in. These documents should be collected and compared after the AI output is shared. Where the AI and the independent reads diverge, that divergence is the most strategically valuable information in the room.
02. Appoint a structured dissenter on every major decision
Sudakov and Furr call this 'intentional dissent.' I call it the one practice that separates boards that make decisions from boards that ratify them. Assign a senior leader — rotating by decision — whose explicit job is to argue against the AI-recommended conclusion. Not to be contrarian. To represent the perceptual ground the model couldn't cover: the weak signals, the adjacent patterns, the local context. Make this role a formal agenda item, not a cultural aspiration.
03. Redesign your leadership development to include judgment preservation
Every AI tool your executives use should come with an explicit protocol for when not to use it. Not because the tool is wrong, but because the capacity to decide without it is what makes the tool useful. AI Vidya's executive programmes are building this into the CXO Leadership AI curriculum for exactly this reason: the skill you need to maintain is independent judgment, and maintaining it requires deliberate practice — not just better prompting.
04. Measure the diversity of your strategic outputs, not just their quality
If your strategy team's quarterly recommendations are becoming more similar year over year — covering the same market segments, reaching the same competitive conclusions — you are watching the Perceived Control Gap in real time. Audit the range of your team's strategic thinking the way you audit the range of your financial assumptions. Compression in either is a risk signal.
The era of AI-augmented strategy is not coming. It is here. The question is not whether your executives will use AI. They already do. The question is whether they will retain the capacities that make them executives rather than editors — the ability to notice what the model missed, and to hold a view it cannot generate.
228 evaluators in a Harvard lab found out the hard way that intelligence and judgment are not the same thing. AI provides the former. Only human experience, deliberate practice, and institutional structure can preserve the latter. The executives who understand this distinction are the ones worth following into an uncertain decade.
Research basis: Sudakov, L. & Furr, N. (2026). "AI Is Undermining Leaders' Judgment. Here's What to Do About It." Harvard Business Review, August 19, 2026. Supporting data: Human Clarity Institute (2026). "How AI Changes Decision-Making." Additional context: Deloitte Insights (2026). "Decision-Making with AI."