I first encountered AI in 1986 and worked with early versions in the 1990s. Even then, it was obvious that human and artificial intelligence needed to work together. It’s the same message I shared decades later in my TEDx talk.

That idea feels even more urgent now. AI learns from the environments we create, and leadership sets the tone. We talk a lot about training AI, but there is an uncomfortable truth: AI is training on us.

AI learns from behavior, not policy

Most organizations still act as though AI pays attention to instructions. It doesn’t. At its core, AI is a mathematical system that learns patterns from data and predicts what is likely to come next, whether that’s a word, a decision, a workflow step or a risk score.

And like humans, it learns from repetition, not rhetoric. ‘Do as I say, not as I do’ never worked on humans. It won’t work on AI.

Your AI doesn’t care about your mission statement. It’s paying attention to how you treat the intern when a deadline is missed.

AI doesn’t watch behavior like a human does. It learns through the digital traces we leave – emails, meeting transcripts and workflow trails.

Across four decades of transformation work, I’ve seen the same pattern: People don’t follow policies, they follow behavior. And now AI does, too. It notices:

• What gets rewarded

• What gets ignored

• The shortcuts and the side steps, even the ones leaders assume no-one sees

If deadlines are rushed, the workflow data shows it. If only certain people speak, the transcripts show it. If leaders quietly fix issues instead of raising them, AI learns that silence is safer than transparency.

AI inherits our gaps

Analysts across Forbes, Harvard Business Review and MIT Sloan highlight that most AI failures stem from human and organizational issues – unclear decisions, inconsistent processes and cultural blind spots – rather than technical limitations. In other words, AI struggles wherever leadership struggles.

Culture is the curriculum your AI studies. AI trains on your data, and your data reflects your culture – your incentives, pressure points, exceptions and defaults.

If leaders consistently prioritize speed over scrutiny, AI will learn shortcuts before it learns responsibility. If important decisions happen out of sight, AI learns that transparency is optional.

Your culture becomes AI’s curriculum. Your behavior becomes its baseline. And like any eager apprentice, AI imitates the environment it grows up in.

I’ve watched leadership teams realize this in real time: that their systems weren’t misbehaving at all. They were behaving exactly as they had been taught.

Inclusion and psychological safety are now risk controls. In many organizations, inclusion is still framed as a cultural initiative. It is that, but in an AI-enabled world, it is also a risk control.

AI inherits the worldview of the people who shape it. If that worldview comes from a leadership team that all think the same way, avoid conflict or make decisions in a hurry, the system will quietly inherit those blind spots.

AI amplifies whatever patterns dominate the room. But psychological safety improves AI outcomes. Harvard Business Review highlights that psychological safety helps teams surface ideas and concerns more openly, which supports better thinking and stronger decision‑making.

In an AI context, early is everything. The sooner a risk is voiced, the sooner it can be corrected, before it becomes embedded and scaled across the organization.

I see this every time I work with teams: When people feel safe to challenge decisions, the AI work improves immediately.

Similarly, cognitive diversity strengthens AI governance. Neurodiverse thinkers often excel at pattern recognition, anomaly detection and sensing ethical tension – exactly the skills AI governance needs.

McKinsey’s long-term research shows that diverse leadership teams outperform on innovation and risk management. The same is true here. Better questions. Better systems.

Inclusion is not a courtesy. It is a control mechanism for responsible AI. When leaders contradict their values, AI learns the contradiction.

Most CEOs genuinely value fairness, transparency and integrity. But AI does not learn from aspiration. It learns from patterns.

• If leaders talk about transparency but make key decisions behind closed doors, AI learns secrecy.

• If leaders champion fairness but rely on the same voices every time, AI encodes hierarchy.

• If wellbeing is celebrated but burnout is rewarded, AI internalizes overextension.

I’ve watched leaders get a shock when their systems reflect behaviors they didn’t think they were teaching. But AI pays attention to the gap between what leaders say and what leaders do.

This is where ethical drift begins.

Being good parents to AI

Mo Gawdat talks about the need to be “good parents” to AI. He’s right. We’re already parenting our systems, whether we meant to or not.

And like any developing child, AI studies the environment more than the instructions.

• It sees how leaders behave under pressure.

• It sees who gets listened to and who doesn’t.

• It sees whether rules are actually rules or guidelines that bend for some and not others.

If AI is the apprentice, the organization’s culture is the workshop – and leadership is the chief craftsperson. The system reflects the environment it grows up in.

This is the philosophy behind my rAIse IT right approach: AI mirrors the culture it’s raised in, not the one leaders wish they had.

The human leadership skills that matter now

Leaders do not need technical mastery to guide AI responsibly. They need to recognize that their behavior is now part of the system.

These leadership skills matter most:

• Ethical clarity: Define what ethical behavior looks like. AI can’t learn standards that leaders can’t articulate.

• Consistency: AI trains on patterns. When humans are inconsistent, the system becomes unpredictable.

• Pattern awareness: Notice what your culture does, not what it claims to value.

• Courageous communication: People need to feel free to ask hard questions. Silence is dangerous when training algorithms.

• Transparency: Closed-door decision-making hides bias. Transparent habits build trustworthy systems.

• Curiosity: Seek anomalies. Ask better questions. Curiosity is governance.

• Cultural humility: Assume blind spots exist. Invite perspectives that reveal them. AI gets better when humans do.

Questions every CEO should be asking now

These are not technical questions. They are leadership questions.

• What behaviors is our AI learning from us?

• Where does bias enter our decisions or systems?

• Who is missing from our AI conversations and why?

• Do people feel safe raising concerns about AI-driven outcomes?

• Are we modeling clarity or confusion?

• Do our lived behaviors match our stated values?

The leadership moment we are in

I’ve spent my career watching technology evolve, but this is the first time I’ve seen it mirror us so directly. AI will change how we work. But its impact will depend far less on algorithms than on the humans raising it.

You wouldn’t hand your teenager the car keys, point them at the freeway and hope for the best. Yet many organizations are doing exactly that with AI: deploying fast, training little and assuming the system will figure out the rules.

We do not inherit the AI future. We raise it. And the way you lead is the way your AI will behave.

Opinions expressed by The CEO Magazine contributors are their own.