Your Emotions Are Doing Calculations AI Can't — And That Makes Them a Leadership Superpower
The people building AI are discovering something that changes how we should think about leadership.
Ilya Sutskever — co-founder of OpenAI, now leading Safe Superintelligence — spent part of his November 2025 conversation with podcaster Dwarkesh Patel on a question that has nothing to do with code: why do humans generalize so much better than AI models?
His answer points directly at emotions.
Sutskever's argument is this: humans have a built-in value function. It's the internal guidance system that tells you, in real time, whether an intermediate step is good or bad — whether you're moving toward something that matters or away from it. This function makes human learning extraordinarily efficient. A teenager learns to drive in roughly ten hours. An AI model can be trained on every competitive programming problem ever written and still fail to generalize that knowledge to an adjacent problem it hasn't seen.
The difference, Sutskever argues, is that in humans this value function is shaped by emotions.
Not managed by emotions. Not corrupted by emotions. Shaped by them.
This is a significant claim from one of the most credible researchers in AI. And it has direct implications for how you lead.
What Business Culture Got Wrong About Emotions
For decades, professional culture has treated emotional signals as noise to be managed — a liability in the boardroom, a distraction from clear thinking, something to be acknowledged and then set aside in favor of data and analysis.
Leaders were rewarded for projecting calm neutrality. Emotional reactions were reframed as bias. The implicit message was: the less your emotions influence your decisions, the better those decisions will be.
Sutskever's framing inverts this entirely.
If emotions are your value function — the system that evaluates whether a direction is worth pursuing — then suppressing them doesn't make you more rational. It makes you less calibrated. You're operating with a degraded guidance system while telling yourself you're being rigorous.
This is not an argument for leading from unexamined impulse. It's an argument for developing enough self-awareness to read your own emotional signals accurately — to distinguish between fear that's telling you something real and anxiety that's just noise, between instinct grounded in pattern recognition and reaction rooted in ego.
That distinction is exactly what AI executive coaching is built to develop.
The Social Intelligence Problem AI Can't Solve
There's a second piece of Sutskever's argument worth sitting with.
He describes how evolution encoded complex social desires in humans — the need for positive standing, the intuitions that govern trust, belonging, and reputation — and admits, genuinely, that he doesn't understand how it did it. Because these aren't low-level signals like hunger or smell. The brain has to perform enormous, sophisticated processing to understand what's going on socially in any given moment.
And somehow evolution said: that's what you should care about.
The result is that humans arrive with social intelligence baked in at a level AI cannot approach. The ability to read a room — to sense when trust is eroding, when someone's agreement is performative rather than genuine, when a team is anxious beneath its stated confidence — requires exactly the kind of high-level social processing that remains AI's deepest unsolved problem.
For leaders, this matters practically. The most consequential moments in organizational life are social moments. The conversation where you have to deliver difficult feedback without destroying a relationship. The team meeting where you have to hold uncertainty without spreading anxiety. The moment you sense that your organization's stated strategy and its actual behavior have quietly diverged.
No tool reads those moments for you. No platform tells you what to do in them. They require presence, attunement, and the kind of social intelligence that took evolution millions of years to build.
This is the territory where leadership coaching operates — and where it will always operate, regardless of how capable AI becomes.
What This Means for How You Lead Right Now
The leaders I work with are navigating real AI pressure — from boards, from vendors, from the general cultural anxiety about being left behind. The instinct, understandably, is to reach for tools. To build AI fluency. To automate what can be automated and move faster.
All of that is right. AI fluency matters and the leaders who develop it will have a genuine advantage.
But the Sutskever insight points to something that gets lost in that conversation:
The capabilities that make you irreplaceable as a leader are not the ones that compete with AI. They're the ones that AI is actively trying to understand and cannot yet replicate.
Your emotional attunement. Your social intelligence. Your capacity to hold a team through uncertainty without pretending the uncertainty isn't real. Your ability to make a judgment call in a situation that has no precedent and live with the consequences.
These are not soft skills. In Sutskever's framing, they are the most computationally sophisticated capabilities that exist. Evolution spent millions of years building them into you. The best AI researchers in the world are studying them because they can't figure out how to engineer them.
The question worth sitting with is not whether you have these capabilities. You do. The question is whether you've developed them deliberately — or whether you've spent twenty years in professional environments that quietly trained you to discount them.
The Work That Actually Matters
David Deming, Dean of Harvard College and the leading economist studying AI's impact on the labor market, has found consistently that as AI automates more cognitive tasks, the premium on human social and emotional skills accelerates. His December 2025 interview with Time put it plainly: adaptability, teamwork, and social intelligence are becoming more valuable in the AI era — not less — because they can't be automated.
Sutskever and Deming are arriving at the same place from different directions. One is building AI and discovering what it can't do. The other is studying the labor market and measuring what it rewards.
Both are pointing at the same set of capabilities.
The leaders who grow through this moment will be the ones who took that signal seriously — who invested in developing the emotional intelligence, social attunement, and human presence that the age of AI is making more valuable, not less.
That development doesn't happen through a platform. It happens through the kind of sustained, honest, human conversation that has always been at the center of great coaching.
If you're ready for that conversation, book a complimentary 30-minute call.
Michael Rolph is an executive coach and leadership advisor with 25 years of experience across technology, government, nonprofit, and international organizations. He works with a small number of senior leaders at a time — globally via video, with in-person availability in the SF Bay Area and Taipei. He relocates to Rabat, Morocco in August 2026.