On August 24, 2001, an Air Transat Airbus A330 took off from Toronto bound for Lisbon. Halfway across the Atlantic, at over ten thousand meters, it ran out of fuel. A leak that maintenance had missed, and that the crew themselves, for a while, read as a faulty sensor. First one engine went out. Then the other.
An airliner carrying 306 people suddenly became the largest glider in the world.
Captain Robert Piché and first officer Dirk de Jager faced something almost no commercial pilot trains for in earnest: flying without engines. With no thrust pushing the aircraft forward, all they had left was altitude, wind, and what they knew. A small emergency turbine, driven by the airflow, gave them just enough to control the aircraft. For about nineteen minutes they glided without power, covering roughly 120 kilometers, until they reached Lajes Air Base, on Terceira Island in the Azores. They landed. All 306 survived.
It remains, to this day, the longest glided landing by a commercial airliner in aviation history.
There’s a certain familiarity in what’s happening in design right now.
What the Engines Did for Us
A modern aircraft flies almost by itself. The autopilot handles the route, the altitude, the descent, most of the flight hours. The system produces the flight. The pilot is there, but for most of the time isn’t piloting in the old sense of the word. They’re supervising.
And then there are the nineteen minutes.
A pilot’s value isn’t measured in the hours when the autopilot flies well. It’s measured in the moment the engines go out over the Atlantic and someone has to decide, with the information they have and the time they don’t, where the plane goes. That moment isn’t produced. It’s judged.
For years, a designer’s work looked more like the autopilot than like the nineteen minutes.
AI Made What Was Visible Cheap
A designer could show their value through what they produced. Wireframes, interfaces, prototypes, components, variations, documentation. Things you could see, present in a meeting, that gave the reassuring sense of work done.
Today, that material comes out in minutes. I ask for an onboarding flow for some product and get five convincing flows before I finish my coffee. I ask for twenty landing page variations and I have twenty. I ask for documentation and it appears.
This is liberating and, at the same time, uncomfortable. Liberating, because the mechanical part stopped eating up time. Uncomfortable, because if a designer’s value was mainly in producing the artifact, that premise changed overnight.
From a Production Problem to a Judgment Problem
This is where the shift is, and it’s worth saying simply.
Design is stopping being a production problem and becoming a judgment problem.
When generating a solution was expensive and slow, knowing how to produce well and fast was, on its own, an advantage. Now that solutions are abundant, that advantage evaporates. Before, the hard part was generating good alternatives. Now that they’re nearly infinite, the hard part is choosing among them.
Chess players got used to this earlier. Stockfish is immensely stronger than any human, and yet the two have learned to coexist well. The machine didn’t push people out of the game, it became the best coach that ever existed. Today it’s easier for a human to prepare and improve, using the machine as a sparring partner. But at game time, human against human or human against machine, the judgment is still human: which move fits this position, against this opponent, at this moment, and when to follow the machine or break from it.
And there’s a part of this that applies even more to design. Judgment isn’t only about what the machine produces, it’s also about what we give it. Which problem is worth solving, what gets left out, what gets done and what gets decided against. Judging what the machine produces is only part of the work. The harder, quieter part is the intent behind deciding what to ask it and what to leave out. And that part can’t be delegated.
When ideas become infinite, judgment becomes the problem.
The Designer Who Only Asks Questions Isn’t Enough Either
Andy Budd wrote two articles that, read together, describe well the tension designers live in.
In the first, “The real reason why executives don’t want you involved,” from 2023, he explains why executives leave designers out of decisions. The answer he gives is uncomfortable, because it isn’t contempt or incompetence. It’s pragmatism. Large organizations have a problem with decisions: every new person at the table brings a new perspective, reopens discussions, and invites even more voices, until the decision stalls. To be able to move forward, leaders do the opposite of what would be ideal. They keep the decision circle as small as possible, shorten the time to decision as much as they can, and say no to anything that threatens to reopen the debate.
That’s also why they prefer outputs to outcomes. An output, a concrete deliverable, can be agreed on quickly by a small group before more people derail the process. And the designer, who tends to ask for more research, more stakeholder input, and the exploration of multiple solutions, becomes exactly the person who reopens what the organization is desperate to close. Budd is honest about this: in an ideal world, the designer would be right. But consensus wears thin over time, and the organization ends up choosing to close fast instead of deciding well.
In the second, “Why designers sound negative, and why that’s a good thing” from 2025, Budd rehabilitates that same behavior. What looks like negativity is, often, anticipation: the ability to imagine what will go wrong before it does. He calls it defensive pessimism and notes that pilots and surgeons use it too. It isn’t a designer’s flaw. It’s insurance against disaster.
I agree with both articles. But I’d add a criticism to them that feels important, especially now.
Asking questions isn’t enough. A designer who only finds problems becomes, themselves, a bottleneck. “Have we thought about case X?” is easy to say and moves nothing.
The valuable designer doesn’t stop there. Instead they say something like: there are three risks we’re ignoring; two are acceptable, one could genuinely affect the outcome, and here’s how I propose we solve it without stopping the project. The distance between pointing out a risk and steering a decision is enormous. It’s the difference between criticism and judgment, and it’s one of the hardest things to learn.
How Do You Learn to Land Without Engines?
And here comes the part that worries me most, because it has no easy answer.
Piché was able to glide that plane because of thousands of hours. Years of decisions, strange situations, small mistakes seen up close. His judgment didn’t come from a manual. It came from exposure to consequences.
Judgment in design is earned the same way. You don’t learn systems thinking by reading a definition. You learn it by watching a seemingly small decision blow up somewhere else in the product six months later. You don’t learn to handle stakeholders in a prompt. You don’t gain product judgment by asking AI for twenty solutions.
You earn it by doing. And it was precisely that doing, the groundwork, the repetitive production, that AI is now absorbing.
The problem isn’t only juniors’, though it’s more obvious in them. It belongs to whoever is mid-career, trying to build a judgment they don’t yet have. And, more uncomfortably, it belongs to whoever is already senior, because much of that seniority was certified in an era when producing well still counted as proof of competence. Is that seniority worth the same now? Part of what was learned was craft, which the machine has taken over. Part was judgment, which remains irreplaceable. It isn’t always easy to say where one ends and the other begins.
If AI removes the hours through which knowledge and experience were always formed, who’s accumulating Piché’s hours?
The Danger of Looking Competent
There’s also a new trap. AI doesn’t just make good designers faster. It makes weak ones faster too. A designer without judgment can now produce more weak work, faster and better-looking than ever before.
This is more dangerous than it sounds, because apparent competence and real competence have drifted apart. The designer’s role increasingly resembles an editor’s. They no longer create everything by hand. They frame the problem, give direction, generate, evaluate, reject, combine, validate, decide. The amount of manual creation goes down. The responsibility for the decision doesn’t go down, and may even go up, because it has never been easier to produce a solution that looks right without being right.
What’s Left to Prove
For years designers asked for the same thing: bring us into the conversation earlier. Don’t leave us at the end, just painting decisions that were already made.
AI will give them that earlier seat, but with a condition. They’ll have to prove why they deserve to be there. And the answer can no longer be because they know Figma, or because they produce better interfaces, or even because they “represent the user.” The machine does a lot of that now.
The answer that’s left is the only one AI doesn’t replace: because they make decisions better.
A designer whose main contribution is turning other people’s decisions into polished screens faces, with AI, a real threat to their place. But for those who help the organization decide better, that same AI may be the biggest lever the profession has ever had.
The autopilot flies fine in good weather. Landing when the weather turns is on you.
Sources
The Air Transat Flight 236, from August 24, 2001, is documented in detail: Air Transat Flight 236 (Wikipedia), Simple Flying and an interview with the captain in The Globe and Mail. It remains the longest glided landing by a commercial airliner.
The two Andy Budd articles I discuss: “The real reason why executives don’t want you involved” (2023) and “Why designers sound negative, and why that’s a good thing” (2025).