Are You Flying, Or Are You Being Flown?

Are You Flying, Or Are You Being Flown?

What a cockpit disaster and three CEO memos reveal about who is actually making decisions in your organisation

Do you remember when those CEO memos started dropping last year? Shopify’s Tobi Lutke went first. Prove AI cannot do the job before you ask me for headcount. Then Duolingo’s Luis von Ahn: AI-first, phase out the contractors. Then Klarna’s Sebastian Siemiatkowski said the quiet part out loud: AI was already doing the work of 700 customer service agents.

If you were reading them at the time, you could see what was happening. They were one-upping each other. Each memo bolder than the last, each CEO racing to describe the workplace of the future as if getting there first was the point. Automate first. Justify the human second.

But the customers hated it. Duolingo’s users hated it. On the company’s TikTok, under a video of a baby owl asking for a cookie, the top comment read: “mama may I have real people running the company.” Von Ahn spent months walking it back. Klarna’s reputation fell off a cliff. Customer satisfaction dropped 22%, and their CEO admitted they had focused too much on efficiency and cost. If your users hate something, they will leave you.

Those CEOs were not necessarily wrong. Maybe it was too much of a knee-jerk reaction, from both sides. But the memos raised a question none of them answered: who decides what the workplace of the future looks like? Nobody knows. We are all playing a game where the rules have not been written yet.
I am no aeronautical expert, but in researching how organisations lose control of their own decisions, one case study kept coming up. In 2009, an Airbus A330 on the overnight from Rio to Paris went down over the Atlantic. 228 people were on board, cruising at 38,000 feet, in the dark. Every system on the aircraft was working. The plane fell out of the sky anyway.

The pitot tubes, small sensors that measure airspeed, iced over. The autopilot did what it was designed to do: disconnected and handed control to the pilots. The problem was that the pilots had been flying with the autopilot for so long they no longer had the feel for flying without it. What happened next took three minutes and thirty seconds. One of the co-pilots pulled back on his control stick. In a stall, that is the worst thing you can do. He needed to push forward. He pulled back.
The stall warning sounded 75 times before the aircraft hit the water. Seventy-five. Not once did any of the three pilots say the word “stall.” Gladwell made famous the idea that 10,000 hours makes you an expert. Those pilots had that. It did not help. Reading the investigation reports, what I see is not a failure of skill. The aircraft was telling them exactly what was wrong. What they had lost was the practice of listening.

Every system on that aircraft worked. What failed was the thing that happens when automation stops: a human takes over. That capacity had quietly gone. Flight by flight, year by year. I have a word for this. I call it algorithmic drift. We go through life optimising and optimising until we blindly believe we are in control of what we are doing. You watch whatever Netflix recommends. You take the first route Google Maps suggests. You book a 30-minute meeting when 17 minutes might do the job better. You become blind to it. You allow the algorithm to do the work, and you are not entirely sure you set the right boundaries. That is algorithmic drift. It happens in your personal life and it happens in your company.

What I call the opposite of algorithmic drift is design. Design is taking control of a situation with your own judgment. Defining what should stay human, what should be handled by AI, and when you need a combination. Because ultimately, we are responsible for the whole system. Not the AI. We are. Perhaps Lutke was ahead of his time at Shopify. He asked the right question: what is AI better than humans at? But what he also did was provide a framework for whether a human should still own a specific job, even if AI could do it at reduced time and cost.

Almost every leader I have spoken to who gets this right has been through something messy first. They felt the pain. And what they learned is that where the stakes are real, you slow the process down. You even break it if you have to. The thing they put back in is friction. Old school. Go through the list. Check every item. Not because they are afraid of AI, but because they are methodical. They are drawing on years of experience and wisdom. Move fast with AI without human judgment and you will break something in a big way. You will have fallen straight into the trap of algorithmic drift without realising it.

The last words on the cockpit voice recorder belong to the co-pilot who had been at the controls when the pitot tubes iced over. Three and a half minutes into a fall he could not stop, the Atlantic coming up at 10,000 feet per minute. He said: “But what’s happening?” He was sitting in front of every instrument Airbus makes. Trained to fly that aircraft. And he did not know what was going on.

If you cannot answer that question about your own organisation right now, about where the decisions are actually being made and by whom, then you are not flying the plane. Somebody, or something, else is