The Hockey-Stick Curse

You've probably seen this line somewhere. On the day Microsoft bought Nokia's phone business, the CEO stood on stage, eyes welling up, and said: "We didn't do anything wrong, but somehow, we lost."

It lands as both tragic and ironic: the company has just been sold off, and the person who held the most power in it can't name a single thing he got wrong?

But here's the buzzkill that has to come first: Stephen Elop never said this.

The full video of that 2013 event is public. The line isn't in it, and nobody cried. The "quote" only surfaced in 2016 — a full three years after the acquisition — out of a viral essay credited to one Ziyad Jawabra, and from there it spread across LinkedIn and a thousand leadership-inspo accounts. Some versions even embellish it, claiming "the entire management team wept openly." A Microsoft engineer, Clemens Vasters, went back and rewatched the footage. His verdict was one word: nonsense.

So why has a fabricated line traveled this far, for this long? Because it understands us too well. It says out loud the lie that every CEO who has driven a company into the ground most wants to believe: I did my best, I didn't do anything wrong — the times left me behind.

And the truth is the exact opposite. What Elop actually wrote — and what the BBC verified as real — was the 2011 internal memo later known as "Burning Platform." Its tone is the precise inverse of "we didn't do anything wrong." It's a confession:

That is the real core of the Nokia story. It didn't lose because it failed to see. It saw — and at the moment it should have jumped, it chose to stay on the platform that was scorching its feet but still printing money.

This piece is about that curse, the one that keeps replaying. It has a name: the hockey stick. And right now it's putting on an AI mask and running again.

To see why this ruler matters so much, you first have to look at one cold curve — it spells out exactly what "not hurting" costs you.

In 2018, three senior McKinsey partners wrote a book called Strategy Beyond the Hockey Stick. They skipped the grand theories and did the grunt work instead: they lined up several thousand of the world's largest companies, ranked them by a metric called "economic profit," and looked at the shape of the line. What they drew made everyone uncomfortable.

What stings more is the lack of mobility. A company stuck in the middle has roughly this chance of climbing into that top 20% over a decade:

What does that mean? It means "playing it safe" is an illusion. A company grows 5% a year, the financials are clean, the board is happy, everything looks steady. But on this curve it hasn't actually moved, because rivals are growing too and the whole industry's water level is rising. It looks like standing still; it's actually being washed slowly downstream. Safe isn't safe. Safe is slow-motion elimination — it just arrives slowly, slowly enough that every quarter you think "fine, let's wait and see."

Amazon's Bezos has a colder name for this state: Day 2.

He insists a company must live forever in Day 1, precisely because Day 2 has no comfortable in-between to linger in. It has exactly one destination.

So how do you climb? The book has one answer: big moves. Not trimming costs or tidying a process, but a handful of moves so large they exceed management's own intuition — bold M&A, yanking resources out of the old business and hurling them at the new, investing capital well above the industry average. The book's math shows that companies making enough of these moves lift that 8% turnaround probability significantly.

Here's the question: if the data is this clear and big moves work this well, why do the overwhelming majority of companies simply not make them? The most valuable sentence in the book answers it. What blocks the way is never failing to understand. It's what the authors call "the social side of strategy." Put plainly: it's human nature in the meeting room. Nobody wants to be the first to float the scary big number. Nobody wants to vouch for a possibly-doomed bet in front of the board. Everyone leans toward setting this year's target a little more conservative, a little prettier, a little safer.

And so, year after year, the forecast management draws is always the same shape: a flat line in front of you, then a sudden upturn at some point in the future, bent into a hockey stick. That upturned tail carries the full hope of everyone who doesn't want to take a risk now. And the tail almost never delivers.

If this still sounds like consultants' chalk-talk, let's look at a few companies that actually died.

They share one frightening trait: not one of them died from failing to see. Quite the opposite — they were often the first in the industry to see the future, sometimes the ones who built it with their own hands. They died because, at the moment to go all-in, they chose to protect the machine still making them money.

Lay these five side by side and the pattern is frighteningly clear. Kodak built the camera. Nokia built the touchscreen. BlackBerry and Intel watched the iPhone being born from arm's length. Every one of them saw it. They died because, at the moment to throw cash, talent, and the whole organization behind it, they took the path that didn't hurt: protect the business still making money, and draw the board another upturned hockey stick.

But if the story stopped here, it would curdle into a dangerous platitude. See it and dare to bet, and you'll always win? No.

So "it hurts, it's irreversible" was never a badge of honor. Hurting is only the necessary condition of a real bet, never the sufficient one. A big move that truly counts both dares to hurt and rests on a judgment that holds up. Courage to bet the house with the direction wrong dies faster and uglier than anyone. Hold onto this dividing line, because it's the one — and the most important — difference between this piece and the cheap slogan that "bravery wins."

History done, back to the present. In this round of AI, who's actually betting?

First, clear out the noise. For two years you've seen astronomical numbers daily, making it sound like the whole world is going all-in. But a large share of those numbers exist to dress up the books, not because real cash went out the door. Run the "does it hurt" ruler over them, and most get crossed off:

There's also something with a familiar smell to it, called circular trade: A invests in B, B turns around and spends that money buying A's products. Both sides' revenue and investment numbers look great, but outside the circle there's no net new real demand.

So that string of trillion-dollar headlines mostly deserves a question mark. They look more like a circle buying from itself to prop up valuations than evidence that end demand is really that large. Cross them off. What's left is the real thing — and the real evidence wears a very plain face: cash that hurts.

And not just the clouds. JPMorgan — a bank — has its CEO saying outright it spends roughly $2 billion a year specifically on AI while saving a comparable amount; its internal AI suite reached 200,000 employees in eight months, and now about 250,000 can use it, half of them daily. Musk's xAI stood up 100,000 top-end GPUs from bare ground in just 122 days, then doubled it in another 92.

These moves share one thing — exactly the litmus test: they hurt, and they can't be undone. Putting cash flow on the line, taking on leverage, pushing a product onto 250,000 people, racing for compute at the speed of construction — all of these are commitments you can't claw back, where being wrong carries a real price. This is what McKinsey's big move looks like: it has to hurt, because the ones that don't aren't big moves at all.

One thing to plant here, so you don't misread it: the clouds are doing the "burning-cash kind of hurt" — they're pouring the foundation for the whole AI track, so their bet naturally sits in capex. But most companies aren't pouring foundations; they're using AI. When it's your turn, the bet that hurts lives somewhere completely different. More on that in Chapter IV.

So what's the contrast? It's the overwhelming majority of enterprises. Run the same ruler over them and most haven't even reached the "it hurts" step:

See the difference? That small group at the top is betting the house in a way that hurts. The rest haven't reached "hurts" at all — they approve a painless budget, buy a few underused licenses, touch neither cash flow nor the organization, with risk exposure near zero.

And risk exposure near zero tells you one thing: this isn't a bet at all. It's more like buying a lottery ticket, propping it on the mantel, and quietly telling yourself "I'm doing AI too." At least the person scratching a lottery ticket knows they're gambling. The real trouble is the comfort of having spent a little money and feeling you're already at the table — it's the same thing as that hockey stick drawn for too many years.

Here a misunderstanding has to be killed on the spot, or everything goes wrong.

The natural reaction: so approve more budget, buy a few more AI tools, open a few more seats — isn't that a big move? No. And this is exactly the trap nine in ten companies are falling into.

Global AI spend passed several hundred billion long ago, yet as noted, 89% of firms report output hasn't moved. Why? Because nine in ten organizations are substituting "buying tools" for "changing the organization." They assume installing AI equals using AI, that issuing licenses equals changing how work is done.

BCG has a rule of thumb, 10/20/70, that cleanly splits where an AI project's value comes from:

Notice it? This loops right back to Chapter I's "social side of strategy." Changing process means touching vested interests; moving roles means facing people's pushback; changing metrics means tearing up a familiar scorecard. These are all political and psychological costs. Buying tools doesn't hurt, so everyone buys tools; changing the organization hurts, so nobody dares. Companies substitute the painless thing — "I've bought a lot of AI" — for the painful thing — "I must rebuild the organization." A perfect re-run of the hockey-stick curse.

The line planted earlier closes here: hurting actually comes in two kinds.

The first hurts on the balance sheet, the second hurts in personnel; who a company is decides which kind of hurt is theirs.

And for those who get it right, how big is the reward? The same power curve — only this time the AI edition. The ~5% of firms out front in BCG's research got:

That's what the top 20% of the curve looks like, concretely, in the AI era.

Of course, this road is never a sure thing. AOL's corpse is still warm — bet the wrong direction, bet too hard, and you die too. This round of AI does have a bubble; those circular trades are the proof. So the argument here was never "AI definitely pays, go all-in now."

Here someone may push back: I'm a regulated financial institution, holding other people's money — "irreversible and painful" isn't courage for me, it's a license-losing way to die. That's exactly right. And it happens to hit a fact a lot of "transformation" preaching skips: for firms like these, reversibility is itself a capability, not cowardice.

But reversible doesn't mean you get to watch forever. The real third road is neither going all-in nor propping a lottery ticket on the mantel:

Most companies fail not from betting too small, but from "staying in the pilot forever, never letting evidence trigger the decision to double down." A pilot from three years ago that's still the same pilot three years later — that is itself a decision, one that reads "let's keep waiting."

So what is the CEO actually supposed to do? On this, Amazon has a framing clearer than most. In his 2015 shareholder letter, Bezos split decisions into two kinds of doors:

What Bezos worried about was the disease big companies catch: running the heavy one-way-door process on everything — layers of review, nobody daring to call it — ending in slowness, timidity, too little experimentation, invention slowly dying. But the companies in this piece made the opposite, and deadlier, mistake: they treated a decision that was clearly a one-way door as a two-way door they could close and revisit later. Digital imaging for Kodak, smartphones for Nokia and BlackBerry — these were one-way doors the industry walks through and never comes back from. Yet they stood at the threshold and, with the casualness you'd give a two-way door — "let's look, invest a little, no rush" — dragged out the once-a-decade decision until the door shut in their faces.

This loops back to the ruler. Why is a one-way door so hard to walk through? Because walking through hurts, and it can't be undone — which is the very definition of irreversible. The human instinct is to quietly recast the painful one-way door as a painless two-way door, so you can comfortably wait a little longer. That's "the social side of strategy," in a more concrete dress.

The real masters use the two doors the other way around: use two-way doors — the small, reversible bets — to test, and once the data gives an answer, find the nerve to step through the one-way door that hurts.

Put these things together and they're three faces of the same thing: