Anna Drobakha, Global Digital Business & AI Transformation Director at Groupe SEB.
At a recent executive roundtable, I put a simple question on the screen: As AI can do more, what must remain human? It shifted the whole conversation. Most of the executive AI debate still centers on capability: Can AI do this? Increasingly, yes. AI can research, synthesize, draft, model scenarios and absorb work that used to require meaningful executive capacity. The more consequential question is different: Should AI do this, and what happens to us when it does?
There’s a temptation to treat automation as a proxy for maturity. Automate 40% of your work and you look more advanced than someone at 20%. I’m not convinced. The best AI-augmented executive may not be the most automated one. What matters is what we delegate, what we augment, what we keep practicing ourselves and whether our distinctly human capabilities get stronger through working with AI or quietly atrophy. I think about this across four dimensions: head, hands, heart and soul.
Head: Augment Intelligence, Don’t Outsource Judgment
AI is becoming a genuine extension of the executive mind. It helps challenging assumptions, synthesize complexity and surface scenarios we’d otherwise miss. That can sharpen strategic thinking. But there’s a paradox: The more AI participates in our thinking, the more deliberately we need to protect the experiences that build judgment in the first place.
Judgment is built by deciding, owning the consequence and discovering where our assumptions were wrong. This isn’t just intuition; a 2025 study of 666 professionals found that frequent AI reliance is linked to measurably lower critical-thinking performance, mediated by exactly this kind of cognitive offloading. If AI frames every question and recommends every course of action, an executive can get very good at reviewing AI’s thinking while getting less practice producing their own.
The fix isn’t rejecting the tool, it’s keeping the human thinking loop alive. Start with your own thesis. Then ask AI to challenge it before asking for alternatives, and notice when and why you override the model’s view. Strategy, creativity and judgment should be AI-augmented, but they have to stay human-first.
Hands: Delegate More, But Deliberately
The hands are different. Research, analysis, meeting prep, production, repetitive workflows—I expect much more of this to move toward AI augmentation and, increasingly, AI delegation. Agents won’t just help with the work, they’ll do meaningful parts of it themselves.
Even so, maturity isn’t proven by delegating everything technically possible. The better question is: Should I do this myself, with AI as a thinking partner, through AI as an agent or through someone else? Those are four different choices with different consequences. Sometimes automation is the right call. Sometimes it’s collaboration. Sometimes a task should stay with a person because doing it is how they build expertise or context they’ll need later. Knowing the difference is itself becoming a key leadership skill.
Heart: Let AI Support Connection, Not Replace It
The boundary sharpens once we move from execution to relationships. AI can help us prepare for a hard conversation by clarifying another person’s perspective, or helping us find a tone that is clearer and more empathetic. That can make us better leaders. But there’s a real difference between supporting empathy and outsourcing relationship. Trust is built by consistency, courage and how we show up when there’s no clean answer. AI can help us prepare for the conversation, but we still have to enter it ourselves.
As routine communication gets easier to automate, leaders may need to be more deliberate, not less, about practicing the interactions that actually build trust. Heart can be AI-supported, but it must stay deeply human.
Soul: The Human Defines The Loop
Then there’s what I’d call the soul of leadership: purpose, values, ethics, accountability. AI systems can increasingly recommend and even execute, and they can be instructed to follow policies and defined boundaries. But someone still has to decide what those boundaries are: what we’re optimizing for, what we won’t do even when it’s efficient, which trade-offs are acceptable.
I suspect “human in the loop” is going to prove too narrow a concept. The human doesn’t just stay in the loop; the human defines it. Organizations will increasingly configure, instruct and evaluate AI systems against their standards and culture. But people must remain responsible for defining those standards, noticing when systems drift from them and answering for the consequences.
A Different Definition Of AI Maturity
First you understand AI, then you use it, then you learn to work alongside it. Mature AI-augmented leadership asks for something more: knowing what to automate, what to augment, what to challenge and what not to delegate at all. And asking honestly whether your human capabilities are getting stronger as a result.
Time saved isn’t the same as value created. More output isn’t the same as better leadership. If AI hands an executive five or 10 hours back a week, the real question is what happens to those hours: More meetings and output, or better judgment and relationships? The same logic applies to capability itself. If AI keeps us from exercising judgment, creativity or hard conversations, today’s efficiency becomes tomorrow’s de-skilling.
We’ve spent a few years asking what AI can do, then how to use it. The next important question will be how we want humans to think, decide and lead in a world where AI touches nearly everything we do. The executives who prove the most AI-mature in a few years will be the ones who can tell us exactly what they didn’t delegate and why.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?