You Cannot Outsource Leadership to an Algorithm
Right now, in almost every organisation I work with, there is a conversation happening about AI that is being held in the wrong room.
It is being held with the IT team. The Chief Technology Officer. The innovation lead. The vendors. Sometimes a consultant. The conversation is about platforms, integrations, budgets, security, models, and roadmaps. It is being framed as a technology decision.
And the most senior person in the room is quietly relieved that it is.
Here is what I want to put on the table. Because this conversation matters, and it is not happening honestly enough.
AI is now making business decisions inside your organisation. Hiring decisions. Pricing decisions. Customer triage. Performance reviews. Marketing targeting. Resource allocation. Risk scoring.
Some of those decisions are still nominally being made by a human, with the AI as a recommendation engine. Most of those humans are accepting the recommendation. That is a decision being made by AI with a person’s signature on it.
If AI is making business decisions, business leaders own the outcomes.
You can no longer outsource the responsibility to IT and call it strategy.
The defensive routine hiding in plain sight
Chris Argyris spent his career documenting a specific pattern in organisations. He called it a defensive routine.
It is the unconscious behaviour leaders use to avoid the discomfort of issues they do not feel equipped to address. The issue gets relabelled. The responsibility gets handed to someone with different expertise. The leader maintains the appearance of being in charge while the actual decisions get made elsewhere.
This is exactly what is happening with AI in most companies right now.
The leader does not understand the technology, so the conversation gets handed to the technologists. The technologists are experts in the system. They are not the ones accountable for the human outcomes the system produces.
So decisions about who gets hired. Who gets the customer. Who gets the loan. Who gets the promotion. Who gets flagged. Are increasingly being shaped by tools whose behavioural and ethical consequences nobody senior is accountable for.
Leadership culture calls that delegation.
Behavioural psychology calls it the defensive routine that hides the real decision behind a technical conversation nobody senior wants to have.
That is the billion dollar blind spot. And it is growing faster than any blind spot I have seen in two decades of working inside organisations.
AI is a leadership problem dressed as a technology problem. And the humans on the receiving end of these decisions do not have the luxury of pretending otherwise.
What AI does to a culture
Here is the part the technology-first framing keeps people from seeing.
AI does not change the laws of organisational behaviour. It accelerates them.
If your culture has psychological unsafety, AI does not fix it. It scales the silence. The same employees who were not telling you the truth about a struggling team will not tell you the truth about a flawed algorithm.
If your culture has unexamined bias, AI does not surface it. It systematises it. The patterns of who got hired, who got listened to, who got rewarded, are now baked into the model that makes those decisions across the whole organisation.
If your leadership behaviour is unclear, AI does not clarify it. It amplifies the confusion. Teams already unsure of the priorities now have to interpret what the system was optimised for. And whether they are allowed to push back when its output contradicts what they know on the ground.
Amy Edmondson’s research on psychological safety draws a line that is more important now than it has ever been.
The question is no longer only whether your team is safe to challenge each other or challenge the leader.
The question is whether your team is safe to challenge the algorithm.
Most are not.
Because the algorithm has quietly become the new smartest person in the room. It is faster than them. More confident. More invisible. And nobody trained your people to push back on a system that does not feel its feelings hurt when they do.
The new learned helplessness
Martin Seligman gave us the language for what happens when humans repeatedly experience that pushing back is futile. They stop pushing back. The behaviour gets extinguished.
Not because people lack capability. Because the system has trained them to defer.
The behavioural risk inside AI deployment is not the technology. It is the way humans learn to behave around it.
Once a team learns that the AI’s recommendation is going to be accepted by leadership most of the time, they stop bringing the harder, slower, more nuanced view that contradicts it. The contextual knowledge. The ethical hesitation. The ground-floor read. All get filtered out.
Not because the system suppressed them. Because the humans around the system learned that voicing them was not worth the friction.
Leadership culture calls that efficiency.
Behavioural psychology calls it the slow extinction of the very judgment you hired your people to bring.
By the time you notice, the institutional knowledge has gone quiet. And the only voice in the room is the one trained on yesterday’s data.
The strategic question nobody is asking
Strategic management treats decisions about technology as tactical. Which platform. Which vendor. Which model. What does it cost. How fast can we deploy.
AI is not a tactical question. It is a strategic one.
A strategic decision is one that defines, over time, what kind of organisation you are becoming. AI is doing that quietly inside every business that has adopted it. The decisions you let it make. The people you let it shape. The behaviours you let it reinforce. Are not technical choices.
They are statements about what your organisation values and who it considers worth listening to.
Every AI deployment is an organisational design decision. Whether you treat it as one is the only thing that determines whether you govern the outcome or wake up to it.
Respect Equality and the power of the algorithm
Here is the part that should sit hardest with anyone holding senior responsibility.
When you write a strategy, you hold the pen, and the pen is power. I have written about this before. Whose voices make it in. Whose get smoothed away. Whose truth gets buried because including it would be inconvenient.
AI is the same problem at a different scale.
Whose data trained it. Whose patterns are now the model’s idea of normal. Whose voices were never represented in the dataset at all, and are therefore now structurally invisible to the system you are using to make decisions about them.
Real people. Real careers. Real lives being sorted by a system nobody in your boardroom fully understands.
This is the heart of what I call the Respect Equality Model. Respect is not a value you hang on the wall. It is a behaviour under pressure. And there is no greater pressure than the moment you have to interrogate a system that is faster, more confident, and more invisible than any human in the room.
If the leader cannot do that interrogation, the system will quietly make Respect Equality decisions on the organisation’s behalf. And the leader will not know which voices have been erased until the consequences are too expensive to ignore.
The leadership test
The leaders who are succeeding with AI right now are not the most technically literate ones.
They are the ones with the highest behavioural maturity. They ask the questions the technology-first conversation never asks.
What is this system optimising for, and is that what we want.
Who in the organisation is allowed to disagree with its output, and what happens to them when they do.
What human judgment is this replacing, and what did we lose when we replaced it.
Whose voices are not represented in the data, and how are we accounting for that.
What are we accountable for, and what have we quietly delegated to the system.
Those are not technology questions. They are leadership questions. And they are the only questions that determine whether AI compounds the value of your culture or compounds the dysfunction of it at speed.
What this requires
If you are a senior leader, the conversation about AI in your organisation does not belong only with IT. It belongs in your boardroom. Your executive team. Your leadership development programs. The rooms where you decide what kind of company you want to become.
You do not need to be able to explain a neural network.
You do need to be able to govern the decisions being made on your behalf.
That is leadership behaviour. It is not a technical skill. It is a maturity. The willingness to hold responsibility for outcomes you do not fully understand the mechanism of, by being clear about the values, the constraints, and the conditions you require any system in your organisation to operate within.
The technology will keep getting better. The leadership behaviour around it has to keep getting better with it. Otherwise the gap between what your AI can do and what your leadership can govern becomes the most expensive billion dollar blind spot of the decade.
We are living through a moment leadership was made for. AI is not the end of leadership. It is the test of it.
AI does not fix bad leadership. It scales it.
Build the leadership first. The technology amplifies whatever it finds. And the humans working inside your organisation deserve leadership that can hold both.
Hacia Atherton is the author of The Billion Dollar Blind$pot, a three-time Amazon bestseller in Strategic Management and Behavioural Psychology. She works with leaders and organisations to build the behavioural clarity that drives both psychological safety and commercial performance
Ready to find your Blind Spot?
You do not get the culture you want. You get the culture your behavior creates.