My Friend's Boss Is Seeing Someone New

I was elbow-deep in the breakfast dishes when my phone rang. It was a friend. A fairly new salesperson, based in Cape Town, and she sounded lost. Properly lost. Her boss, who sits in Durban, had just emailed the team to announce he’d signed up for a paid version of a frontier AI model. Things, he said, were going to change. For the better!

Then came the follow-up email. New KPIs and sales targets, designed by the model, to be implemented immediately. No nuance. No context. Nothing about location, history, or what the market is actually doing right now. Just a spreadsheet, and whatever the model had absorbed about salespeople in her industry. One assumes. One hopes.

“How do I manage this?” she asked.

Which is a wonderful question, when you think about it. Her manager had just announced his new relationship with an AI, and she was the one left wondering how to manage it. Or him.

I put the plate down.

She wasn’t the first. A few weeks earlier a senior engineer had called from another city with almost the same story. Then someone in finance, who sits on an executive board. Three calls in five weeks, same topic.

With the first two, something had been building. But they’re senior. They know how to push back. She was new. She had the least room to argue, and no clear idea who she’d even be arguing with. That was the call that stopped me.

While much of the world is getting its knickers in a twist about em—dashes and who’s using AI to write what, something far bigger is happening, and we should probably talk about it.

I get the appeal. I’ve watched people become hypnotised by what these models can do. I’m a little hypnotised myself, if I’m honest. But this isn’t about emails anymore. It isn’t about an Excel workbook. Leaders and managers are handing over their decisions, their responsibility and their authority, and the people who report to them are left wondering who they actually work for.

THE AI HAND-OFF

So I went looking. Was this bigger than my three mates?

Yup, it is. It isn’t confined to one part of the world, and it isn’t confined to one industry. And one story in particular has stuck with me.

An attorney in the US told Futurism about a boss whose interest in ChatGPT slowly turned into something closer to devotion. First, staff were told they had to run things past the AI before every meeting. Before they even spoke to him. Then he started using it to decide who to hire, and who to fire.

Here’s the twist. The company was on a shared paid account, which meant his staff could see his chats. So they looked. They watched him ask the AI about them, and kept checking back to find out who was about to be fired and who was in line for promotion.

Read that again. The employees weren’t reporting to their manager. They were reading over his shoulder while he reported to a chatbot.

In the end she quit. The final straw was an AI-built handbook the staff called “The Bible”.

And it isn’t just a handful of rogue bosses. A survey of more than 1,300 US managers found that, among those using AI, around two-thirds had used it to help decide on layoffs. More than one in five said they often let it make the final call, with no human involved at all.

We’ve moved well beyond asking AI to help with the work. For a while, the questions sounded like this:

• Can you help me draft this email?
• Can you summarise this report?
• What are my options?

In the AI hand-off, they sound like this:

• Who should I hire?
• What should our strategy be?
• Who’s underperforming?
• Who should I let go?

Same chat window. Very different questions.

YOU SIGNED UP FOR THIS

To be clear, I’m not blaming AI for any of this.

When I talk to people about it, the conversation almost always drifts to what AI can and can’t do. Is it accurate? Is it biased? Is it good enough yet? Interesting questions. Wrong ones.

Even if AI were brilliant at this, the job would still be yours.

This is a human problem. A leadership problem. A cultural problem. It’s a person deciding to hand the responsibilities they signed up for to anything, or anyone, other than themselves.

Remember my friend’s boss? He signed up to a paid AI model. Good for him. But long before that, he signed up for the job. One of those comes with a monthly fee. The other comes with people.

And the job comes with things no model can carry for you.

• It can’t sit across the table from someone and tell them they’re being let go.
• It can’t be called into a room to explain itself.
• It has no skin in the game. If the strategy fails, it doesn’t lose its business.
• It has never sat in a performance review where the outcome could cost it its own job.

The responsibility stays exactly where it was. One company in the US found that out the hard way. It used ChatGPT to draft its explanation to a government agency of why it had fired an employee. The AI-written reason included the fact that the person had been let go for discussing their pay, which is protected under US labour law. Using ChatGPT didn’t get the company off the hook. The judge’s view, more or less, was that whoever sent it could have read it and taken that part out.

You can hand off the task. You can’t hand off the blame.

Which, if we’re honest, is something leaders have been trying to do for a very long time.

THE AI SAID SO

Take a little trip down memory lane and you’ll hear it.

“It’s company policy.”

“The consultants recommended it.”

“The numbers don’t lie.”

“The committee decided.”

“It’s the market.”

Managers have always had somewhere to put an unpopular decision. Somewhere that wasn’t them.

“The AI said so” is just the newest line on the list. But it might be the best one yet. It’s fast. It’s articulate. It’s confident. It never needs a coffee break, and it can be rolled out across an entire organisation by lunchtime.

And then there’s the big one. It looks neutral.

A consultant had a fee and a reputation to protect. A committee had names in the minutes. You could argue with them, or at least find out who to argue with. AI arrives with no agenda, no favourites, and no face. So when it hands down a decision, that decision feels objective, even though it’s answering the question the manager asked, in the way the manager asked it.

That’s what makes it such a good place to hide.

To be honest, AI hasn’t created a new problem. It’s given a very old one a much better disguise.

YOU’RE STILL THE BOSS

I’m not done with this. It’s far too interesting a trend to leave at the bottom of one post. Two questions in particular keep nagging at me. Does AI create space for better human leadership, or does it quietly reduce the need for it? And what would responsible AI leadership actually look like, on a Tuesday morning, in a real office? I’ll be back with those.

For now, a word to the managers and leaders who’ve just signed up.

Use it. Genuinely. Let it draft, summarise, crunch the numbers, model the options, and even argue with you. Let it take the admin off your plate so you have more time for your people, not less.

But when the decision is about a person, make it yourself. Put your name on it. Be the one they can find when they want to understand it.

Because somewhere on your team, right now, someone is phoning a friend and asking, “How do I manage this?”

They shouldn’t have to. That’s your job.

#NufSed