It Was Never About the Technology

Not long ago I worked with a manager who had a practice. Every time someone brought him something to review, a doc, a deck, an analysis, his first question was the same. "Did you ask AI?" At first people thought he was needling them. Then they understood. He wasn't asking whether they'd cut a corner. He was asking whether they'd pressure-tested the thing before it reached his desk, whether they'd handed it to that tireless skeptic and made it argue back. To him, running your work past AI wasn't cheating, it was adding a layer of hardness to it. The team adjusted and they started showing up with work that had already survived a round of questioning. This was more than a year ago, when almost nobody was working this way.

Compare that to a story a client told me recently. She'd presented some work to the head of her firm, who glanced at it and said, "Oh, you probably used AI for this." She had. But that wasn't the point, and the comment wasn't a compliment. It meant the work was somehow less hers, less real, less earned. She left the room feeling dismissed.

Same four words, more or less. Did you use AI? In one room it means you did your job well. In the other it means you didn't really do it at all.

I keep coming back to that gap, because I think it explains something a lot of organizations are getting wrong about this moment.

It helps to be precise about what AI is. It's not a product you buy, not a feature you switch on, but something economists call a general-purpose technology. The steam engine was one, electricity was one, the PC and the internet were the last two. What sets these apart from ordinary innovations is that they don't improve a single task. They reset how work gets done across every industry at once.

They also share a frustrating trait. The payoff shows up late. When factories first electrified, productivity barely moved for years. The motors were installed, the wiring was run, and not much happened. The economic historian Paul David told this story well. The value was never in the electricity. It was in everything you had to rebuild around it. Factories had been designed around a single central steam engine, machines crowded close to the power source. Electricity let you put a motor on every workbench and lay the floor out around the flow of the work itself. That redesign took a generation, because it wasn't a wiring problem. It was a human one. People had to rethink the job.

That lag, between when the technology arrives and when the value lands, is the most reliable pattern in the whole history of general-purpose technology. The technology shows up. It's necessary. And yet it is nowhere near sufficient. A computer on every desk didn't change how companies worked until people had applications worth running and reasons to use them. An internet connection plugged into every office didn't transform business until services emerged that actually relied on it, things like email and e-commerce and cloud-based tools that made the old workflow obsolete. In each case, having the technology was the price of entry. The value came later, when organizations figured out how to rebuild the work around it.

AI follows the same pattern, with one difference worth naming. Every previous general-purpose technology still required capital to access. You had to buy the mainframe, wire the factory, build the network. With AI, you open a browser. Your competitor opens the same browser the same morning, to the same model, at roughly the same cost. That doesn't change the pattern, it sharpens it. When the technology is free and universal, it stops being a source of advantage entirely, because everyone has it at once. The only thing left to compete on is the slow part: how fast your organization can actually change the way it works. And that can't be bought or rushed, because it runs at the speed of people.

The catch is that because the value lags, the data lags with it. Pull the numbers on most AI investments today and they look underwhelming. Pilots that didn't scale. Licenses nobody fully uses. Productivity charts that haven't moved. A careful executive looks at that and draws the responsible conclusion. This is overhyped, let's wait until it matures. Let's not bet the quarter on a maybe. And for this quarter, they're right. The data supports waiting.

The trouble is that data only ever describes the past. Clay Christensen made this point in an interview I still show my Strategic Management students. Data is available only about things that have already happened, he said, so when we train people to be data-driven and analytical as they look toward the future, we "condemn them to take action when the game is over." The only way to see the future, where by definition there is no data yet, is through a good theory. The companies that move early on a general-purpose technology aren't reading better numbers than everyone else. There are no better numbers. They're operating on a theory of where this goes, and acting on it before the evidence shows up to make the decision feel safe.

By the time the evidence shows up, the game is mostly over, and the gap closes on the people who waited for it.

A theory at the top of an org chart doesn't do anything on its own. It has to move the organization, and organizations don't move because of a memo. They move because of who gets listened to.

Every company has people who set the weather. Not always the most senior name on the chart, but the one whose questions carry, whose interest signals what's worth doing, whose raised eyebrow can quietly kill an idea before it's spoken aloud. When that person treats AI as a serious capability, the rest of the organization leans in. When that person treats it as a gimmick, or worse, as cheating, everyone gets the message and goes quiet.

There's a catch most change efforts walk straight into. You usually can't swap that person out. The weather-setter holds that role for structural and institutional reasons, and those reasons don't move. What you can change is what they believe. Which sounds like a job for persuasion, a good deck and the right statistics. It almost never works that way. People don't reason their way into a new relationship with a tool. They get their hands on it, they use it for something that matters to them, and the conviction follows the using. Behavior first. Belief catches up.

I'm doing exactly this with a firm right now. When we ran the initial assessment, we found most of the employees had already been experimenting with AI, some more than others. What they were waiting for was a signal from the top that it was safe and expected. The signal they'd gotten ran the other way. About a year earlier, one of the senior executives had reacted badly to some people using AI. Whatever the actual circumstances, the message the employees took away was simple. This executive doesn't like AI. So the experimentation went underground, and the business stalled, not for lack of interest but for lack of permission.

So the work didn't start with the staff who were already curious. It started with the executives, and with that one executive most of all. Not a lecture on AI's potential. Getting him to use it himself, on a problem he actually cared about, until he saw the value with his own eyes. Because the day he starts asking "did you ask AI?" and means it the way my old manager meant it, the whole company changes with him.

Which is what those two managers were telling us all along. They aren't two kinds of people. They're the same seat, before and after. The CEO who waved off his employee's work is the executive who hasn't picked it up yet, grading the tool, still treating it as a shortcut someone took instead of the work itself. My old manager is that same role after the conviction lands, asking the same question, but with the opposite meaning. The difference between those two companies isn't the technology. They have the same technology. The difference is whether the person who sets the weather has changed his mind yet.

I learned how expensive that difference can be a long time before any of this.

From 2001 to 2004 I worked at Sony, in the corporate arm, on a team building platform technology meant to connect the company's US business units; Sony Electronics, Sony Pictures, and Sony Music. On paper, Sony should have owned what came next. They had the hardware. They had a music label. They had invented the portable music market two decades earlier with the Walkman, and they'd been in digital portable audio since 1992, almost a decade before the iPod. Every piece you'd need to build what Apple eventually built, Sony already had in the building.

What we ran into, over and over, was the silos. Each business unit saw only its own P&L, its own targets, its own turf. The electronics business worried that a real digital music player would undercut the music business. The music business worried about protecting its content. Nobody at the unit level could see the thing that was obvious from where we sat, which was that the internet was about to make those walls irrelevant, and that whoever connected them first would win. We were trying to build the bridge. The business units didn't want to cross it.

Apple had no such problem. No label to protect, one P&L instead of warring business units, nothing to defend. They assembled the pieces Sony already owned and couldn't put together. Sony didn't lose the portable music market because it lacked the technology. It had the technology first. It lost because it couldn't reorganize itself around what the technology made possible, and the people who set the weather inside each business unit had every reason to keep the walls exactly where they were.

That was more than twenty years ago, and the lesson hasn't aged a day. The technology is never the advantage. Everyone has the same technology. The advantage is whether your organization can absorb it, and absorption comes down to whether the people who set the weather can change their minds before the evidence forces the issue. Because by the time it forces the issue, the window has usually closed.

So the question worth asking inside your own company isn't whether your people are using AI. Plenty of them already are, quietly, waiting to find out whether it's safe to admit. The question is what happens when the work reaches the person who sets the weather. Does that person ask "did you use AI?" the way my old manager did, meaning did you make this harder to argue with before you brought it to me? Or does that person ask it the way the CEO did, meaning this isn't really your work. Same four words. One of them is building a company that absorbs what's coming. The other is building a company that finds out too late.

Break a Pencil,

Michael

P.S. If this piece resonated, share it with a leader in your organization who sets the weather. They might not read it because they need to. They might read it because someone they trust sent it to them. That's how belief starts to change.