Most leaders steering an organization through the AI shift have the same private feeling: that this one is different. The pace is faster, the capability is broader, and the playbook from the last technology cycle doesn’t seem to apply. The feeling is understandable. It is also, in one specific way, wrong.
The tools are new. The shape of the change is not. It has happened repeatedly – in aviation, in professional kitchens, in manufacturing, and in fields as different from each other as medicine and accounting. Each time, a capability arrived that took something expensive and made it cost almost nothing. Each time, the transition that followed moved through the same four phases, in the same order, for the same reasons. Three of those cases are unusually well documented, so they’re the ones worth walking through. Knowing the pattern doesn’t tell a leader what to do next. It tells them where to look – and which instinct to distrust.
The four phases
Every one of these transitions ran the same arc.
It starts with delight. The new capability is intoxicating and almost free. Anyone can do the thing that used to require years of training. The results are spectacular – the good kind and the bad kind, often from the same person on the same day.
Then comes craft. The first failures force a discipline: a standard way of giving instructions. Output gets more predictable. Skill starts to matter again, but now it’s skill at using the method, not raw talent.
Then embodiment. The method isn’t enough on its own, so the environment gets built up around the person – the right information, the right tools, delivered to the right place at the right time. The individual doesn’t get smarter. Their surroundings get smarter.
And finally systems. The environment becomes a coordinated whole, with feedback loops, checkpoints, hard limits, and someone accountable for quality across the entire operation. At this phase the system, not the individual, is what produces reliable results at scale.
That sequence is abstract until you watch it happen. So watch it happen three times.
I can’t take credit for the pattern itself. The engineer and writer John Crider has traced it through these same three domains; what I want to add is what it means for the people steering an organization through it.
Aviation: from barnstormers to air traffic control
In the 1920s, surplus war pilots bought cheap biplanes and flew them into farmers’ fields, selling rides for a dollar. No instruments, no training, no coordination. The barrier to entry was nearly zero, and the results were exactly as spectacular – and as fatal – as you’d expect. That was the delight phase.
The first real improvement wasn’t a better airplane. It was the checklist. After a prototype bomber crashed in 1935 because the crew skipped a step, aviation adopted standardized pre-flight procedures. The aircraft didn’t change. The way humans operated it did, and the safety gain was enormous. (The same insight later reshaped surgery – the operating-room checklist comes from exactly this lineage.) That was craft.
Then came the instrument panel: the artificial horizon, the altimeter, radio navigation, weather reports. Pilots were given the information they needed to make good decisions in conditions where their own senses lied to them. That was embodiment.
Modern aviation is the systems phase. Air traffic control imposes real-time constraints. Radar continuously verifies where every aircraft actually is. Collision-avoidance systems will override the pilot’s own controls if a crash is imminent. No pilot, however skilled, can safely move a full airliner through crowded airspace without that system around them. The system doesn’t replace the pilot’s skill. It makes that skill reliable at scale.
The kitchen and the factory
The same sequence, twice more, compressed.
A restaurant serving two hundred plates on a Friday night does not rely on one gifted cook. It runs on a brigade: standardized stations, a sous chef inspecting every plate before it leaves, timing called by an expediter, the rule that no plate goes out until the whole table’s order is ready. Cooking by feel was the delight. The recipe was craft. The prepped, organized station was embodiment. The brigade is the system – and notice what it does. It doesn’t eliminate the creative cook. It creates the conditions where improvisation is consistently excellent instead of occasionally disastrous.
Manufacturing is the clearest case, and it’s worth following it all the way to the car. Before the factory, a master craftsman could produce a remarkable clock or violin by feel – and no two were ever identical. Standardized, interchangeable parts were craft: output you could finally count on. Then the work moved to the automobile, and the moving assembly line was embodiment – each worker given exactly the parts, tools, and information for one operation, the car coming to them rather than the worker chasing the car. The Toyota Production System and the quality movement that followed were the systems phase: not a better car, but a better system for building every car the same way. W. Edwards Deming’s central finding is the one every leader in this transition should sit with: quality is a property of the system, not the individual. You don’t get reliable quality by hiring better people. You get it by building a better system around the people you have.
Where this leaves you
AI-assisted work is running the same arc. The early free-for-all was the delight. Standardized prompting was the first craft. Giving the tools the right context was embodiment. What’s forming now – the verification steps, the limits, the accountability for what the tools produce – is the systems phase, and we are in the messy middle of it. That middle is always messy. None of these transitions was clean or planned. Aviation didn’t adopt checklists because the idea sounded good in a meeting. It adopted them because people died. Manufacturing didn’t embrace quality systems out of conviction. It did it because competitors who had them were winning.
Which brings the lesson worth keeping. At the systems phase, “just get a better tool” is always the wrong answer. A better airplane doesn’t fix the absence of air traffic control. A more capable AI model doesn’t fix a missing system around it. When the problem is systemic, only a system fixes it – and the instinct to solve it by upgrading the tool is the single most reliable way to stay stuck.
Individual skill still matters. It has in every one of these stories. But in none of them was it the bottleneck once the transition reached this phase. The best pilot can’t land safely at a crowded airport without the system. The most gifted cook can’t serve a Friday night’s two hundred plates without the brigade.
So the question for a leader isn’t “do we have the best tools.” It’s quieter and harder: which phase are we actually in – and are we trying to buy our way out of a problem that only a system will solve? The organizations getting real value from this transition aren’t the ones with the best models. They’re the ones whose leaders recognized which phase they were in, and stopped reaching for a better tool once a better tool was no longer the answer.
This has happened before. That’s the good news. We know how the story goes, and we know what the people who navigated it well were paying attention to. It wasn’t the tool.
The four-phase pattern in this piece is drawn from J.M. Crider’s Harness Engineering for Vibe Coders (2026); the reading of it for leaders is my own.
