If we join companies and leave leaders, what is there to leave when the leader is AI?

There is a phrase every HR executive has used, usually while nodding gravely: ‘People join companies, but they leave leaders’.

We’ve all felt it. The manager who made an otherwise good job feel like slow torture, or the one whose approval we chased longer than we should admit. The phrase is usually traced to Gallup, and to Buckingham and Coffman’s ‘First, Break All the Rules’, and its emotional accuracy is why it has survived a quarter of a century of repetition.  Many of us have, at some point, quit a job to get away from a person rather than a place.

So it is a good line. It is also, on the evidence, only half right.

01 — THE CRACK IN THE ADAGE

Why we leave is rarely just the leader

When Culture Amp’s scientists went looking for the research underneath the saying, they couldn’t find it. Their own data put the immediate manager’s weight on the decision to leave at around a tenth. Real, but still modest. Senior leadership mattered more than double that. And development opportunities (whether there was any possibility to grow), dwarfed everything else. The finding was that in organisations with poor development opportunities, the gap in commitment between people with a good manager and a bad one was actually zero. So when there is no future on offer, the quality of the person offering it stops mattering.

What this shows is that a leader does two jobs at once. Day to day, they set the ‘texture’ of our working life. Gallup still finds that the leader accounts for roughly 70% of the variance in how engaged we are. But the leader is also often the ‘gate’, namely the person standing between us and whatever comes next. The texture is what we feel. The gate is what we leave. A great leader can make a dead-end job bearable; but they cannot, on their own, build a staircase that isn’t there.

02 — WHAT A LEADER ACTUALLY BRINGS

The climate they create, not the tasks they assign

Strip a leader down to their function and we get a task allocator, a performance rater, a signer of leave forms. But that has never been what we mean when we talk about leaders who shaped us. We mean something more like weather: a particular person created a particular kind of climate, and we grew (or didn’t) in it.

Like most of us, I have had the full range.

The one who gave me weather to grow in

I was in my twenties when I got my first great leader. Before that, it had been a run of the uninspiring and, at times, frankly toxic. But this one saw something in me I couldn’t yet see, and kept handing me the room to prove it.  They were warm and encouraging, but exact about what good looked like and what didn’t. When the big promotion came up, I pushed back saying I wasn’t ready. Thankfully, they didn’t listen.

And so I soared. The work was fun, the team was tight, and I gave that organisation everything.  The discretionary kind of effort no contract can compel.

And the ones I’d rather forget

But I’ve had more that weren’t that. Like the one who pitted team members against each other. Or the one who told a long-serving employee he’d been demoted, in front of the son who’d popped in after school. Or the one who transferred a loyal colleague to another team by email. Or the one whose moods had everyone walking on eggshells.

We cannot overstate what that does to a team. Trust thins, people grow wary, and the organisation gets exactly the bare minimum it has earned.

But despite all this, I still believe in the human leader.  Because the personality that can wreck a team is the same one that can lift it, and either way it runs right through the work. Which is precisely why the next question is worth asking properly.

03 — NOW TAKE THE COLOUR OUT

What happens when the leader is a bot

This is no longer hypothetical. In the OECD’s employer survey, some form of algorithmic management (software, increasingly AI, doing the work managers used to do, such as allocating tasks, monitoring, rating and nudging) already reaches around 90% of firms in the United States and nearly 80% across Europe. The manager role is quietly being unbundled. Some of what leaves is the bad weather. But some of it is the weather entirely.

The instinct is to assume this must be worse, and sometimes it genuinely is. The broader research on algorithmic management is a fairly grim catalogue of more surveillance, intensified work, thinner autonomy, greater insecurity, and lower satisfaction. But the honest picture has a twist that is unsettling. In one controlled experiment, workers reported no significant difference between being managed by an AI and by a human. Not on performance, motivation, or even fairness. The bot, it turns out, is not obviously the villain, which is interesting.

And then there is the finding I’m genuinely curious about:

Workers managed by an AI will accept lower pay, and stay just as motivated.

It seems that when we are managed by a person, a low offer is an insult. It arrives inside a relationship, and relationships can be betrayed. But when we are managed by a machine, the same low offer is just a number. Nobody withheld anything from us. Nobody looked us in the eye and decided we were worth less. So apparently we shrug and carry on, and we do not even lose heart, because heart was never in it. To me, something about it feels akin to going to the supermarket and just accepting that the price is the price, no haggling.

04 — THE SETTLEMENT

It’s hard to leave someone who was never there

The old adage says the leader is who we leave. But a leader is also who we stay for, over-deliver for, forgive a bad quarter for, follow to the next company for. The best ones build a climate with so much heart that the place becomes genuinely hard to leave. Leaving feels less like resigning than like leaving home. And a good one also keeps paying out even after we’ve gone. People who leave a leader they respected tend to leave warmly, referring clients, sending talent, sometimes coming back. That goodwill is the return on being known as a person, the one thing a system that only ever processed you can never earn.

Take the person out and some of the ledger can genuinely improve. No favouritism, no moods to read, no boss whose bad day becomes our worse one.  And there is real research that a bad manager measurably damages our health, so this is not nothing. A well-built system can be more consistent and less political than the average human it replaces. If you have ever been managed by someone insecure, you already know the appeal of a manager with no ego to protect.

But the same subtraction removes the thing that made the relationship worth more than its terms. It’s hard to leave a manager who was never a person. And we cannot be inspired by one either. What is left is not a worse relationship, it is not a relationship at all. It is merely a ‘settlement’: clean, legible, transactional, and quietly cheaper, because there is no one there to be generous with us, and therefore no one we feel we owe.

So where does this leave me? After every difficult leader I would still, without hesitation, choose a human over a bot every time. Not because humans are safer, they plainly aren’t. But I would choose the human for one reason, and it happens to be the reason I am here at all: a bot would never have made me take that promotion.

It would have read my hesitation as data: ‘Candidate declines; preference logged; move on’. My leader read it as fear, and refused to honour it. They saw something I couldn’t yet see, and were willing to be wrong about it out loud. That isn’t a calculation, but a kind of faith, and faith is the one thing an algorithm cannot extend, because it only ever reasons from the evidence already in front of it. A machine can be scrupulously fair to who you are. Only a person can be unfair, in your favour, about who you might become.

And yes, the same current runs both ways. The leader who believes in you and the leader who demotes a man in front of his son are drawing on the same human capacity to reach into another person and leave a mark. Take it out and you lose both.  The wound and the lift. That is the honest trade, and I would still make it. I have never once grown in climate control.

So let me name the distinction we have been circling all along. Management is the mechanics: setting the work, tracking it, rating it, keeping it consistent and fair. A machine can do all of that, and maybe better than we did. Leadership is the rest, the part that wasn’t written into the role. It is the leader who put me up for the job I swore I wasn’t ready for. It is a climate with enough heart that leaving felt like leaving home. It is being believed in before the evidence is there. AI can manage, it cannot lead.

Which is why the real question was never whether AI makes a good manager. The question is what we think an employment relationship is for. If it is a mechanism for extracting agreed work at an agreed price, then we have the option to hand it to the machine. But if it is the place where people are seen, inspired, stretched and occasionally changed, then something is being quietly removed while we congratulate ourselves on the efficiency.

People join companies and leave leaders. The unnerving future is not one where the leaders get worse. It is one where there is no one to leave, and no one, either, worth staying for.