Part 3 in The Autopilot Corporation series. Part 1 here. Part 2 here.

It’s brutally hard to start a company. Anyone who’s done it knows. You spend years looking for the thing — the product, the wedge, the customer who actually pays — and most of those years are quiet. Nobody writes about you. The trade press doesn’t know your name. There’s no scorecard, no rating, no allocator with a thesis. There’s only the person on the other end of the transaction, and whether their wallet opens.

That’s the lightning in the bottle. The founders I respect can describe it without flinching. They know who their customer was, what they paid for, and why. Usually in one sentence. Sometimes one word.

Then the company succeeds, and something starts to change.

The drift, again

I wrote about this at the company level in Part 1 and the Part 2 warning signs piece. The short version is that success changes the audience. Once a company is winning, new voices crowd into the room. The board. The analysts. The bond market. The trade press. The activist investor with a slide deck. The HR consultant with a framework. Each one has a metric they want to see move, and each one has just enough leverage that ignoring them feels expensive.

The company stops describing itself by what the customer gets and starts describing itself by category. “We’re a cloud software provider.” “We’re the digital-first bank.” Strategy decks replace product decisions. Innovation labs replace innovation. Boardrooms become country clubs. Budgets become photocopies of last year. The culture turns from finding the customer into managing the optics.

What I called drift in Parts 1 and 2 is what happens inside one company when this pattern compounds. Performative work multiplies. Internal research gets commissioned to justify decisions already made. Compliance offices grow. The company becomes a system optimizing for signals that have nothing to do with the person who originally opened a wallet for it.

Eventually one of two things happens. The company dies. Sears. Kodak. Blockbuster. The 1990s GE. Or someone arrives whose only job is to take the controls back.

The reset CEO

The pattern is well-documented even if it’s rarely named cleanly. Lou Gerstner at IBM in 1993. Steve Jobs returning to Apple in 1997. Alan Mulally at Ford in 2006. Satya Nadella at Microsoft in 2014. They didn’t invent anything new. They remembered what the company had forgotten. Each one walked in, looked around, and asked the question the previous leadership had stopped asking: who is our customer and what do they actually pay us for?

Then they killed the things that didn’t connect to the answer. Gerstner killed the plan to break IBM into hardware silos and bet on services. Jobs killed seventy percent of Apple’s product line in his first week. Mulally sold Jaguar, Land Rover, Volvo, and Aston Martin so Ford could mean one thing again. Nadella put Office on iOS and Android, which his predecessor would have considered heresy, because the customer was already on iOS and Android.

I’d add a recent one. In September 2024, Brian Niccol took over at Starbucks and wrote an open letter that included this line: “there’s a shared sense that we have drifted from our core.” A sitting CEO of one of the most recognized brands in the world used the word drifted. He named the disease in public, in writing, in the same vocabulary I’ve been using in this series. (Coffee Intelligence, 2024.)

The reset CEO move isn’t magic. It works less than half the time. I went through thirteen well-known cases and the hit rate landed somewhere around forty to forty-five percent. Mayer at Yahoo failed. Ron Johnson at JCPenney failed catastrophically. Schultz’s second return to Starbucks was essentially a holding pattern. Iger’s return to Disney has been mixed at best. What separates a real reset from a cosmetic one is whether the new leader made an irreversible structural bet on the customer, or whether they just announced one and trimmed costs.

But when it works, the mechanism is always the same. Strip back. Remember who pays. Ignore the noise.

And what if the whole economy did this at once

This is where Ray Dalio is useful, lightly. Dalio writes about long cycles. His Big Cycle frame describes upwaves where debt and financial wealth grow faster than the real economy, until the gap can’t be sustained and the cost of capital re-prices (Principles for Dealing with the Changing World Order, 2021). I’m borrowing the shape of his argument and leaving the geopolitics on the shelf. The piece of it that matters here is simple. When money is cheap for long enough, incentives drift. When money gets expensive, they snap back.

The Federal Reserve held interest rates at or near zero for roughly twelve of the fourteen years between 2008 and 2022. (Forbes Federal Funds Rate History.) There’s no precedent for that in the postwar period. The ten-year Treasury yield hit a low of 0.52 percent in August 2020. Investment-grade corporate bond spreads compressed to their tightest levels in nearly two decades. (European Central Bank, 2025.) Global venture funding hit $621 billion in 2021, more than ten times the level of a decade prior. (CB Insights, Q4 2021.)

I keep coming back to that fourteen-year window. It wasn’t a moment. It was a generation of capital. An entire cohort of executives spent their formative leadership years inside it. If you ran a public company between 2010 and 2022, your career was built in an environment where the cheapest path to capital didn’t run through the customer. It ran through the allocator.

What Patrick Berzai noticed

A few weeks ago, Patrick Berzai published an essay called Did the customer ever ask for this? He’s running the same diagnostic I’ve been running on companies, but at the level of the whole economy. His core observation is that for most of the cheap-money era, the marginal decision-maker inside large public companies stopped being the person with a wallet and became the institution with a metric. ESG ratings. DEI dashboards. AI strategy slides. Compliance scorecards.

His sharpest line is that this looked like capitalism but functioned closer to Soviet-style central planning. Targets set by allocators. Companies optimizing for the targets. Customers downstream from the whole arrangement. He calls it early-stage Gosplan. I think that’s a touch overstated — Soviet planning eliminated price signals entirely, and ZIRP-era companies still had revenues and customers, they just deprioritized them — but I get what he’s reaching for. The mechanism he’s pointing at is real.

What I find most useful in his framing is the question itself. Did the customer ever ask for this? That’s the same question Gerstner asked at IBM and Nadella asked at Microsoft. He’s just asking it of a whole decade.

Where I’d extend him

There’s a piece Berzai’s essay doesn’t fully name, and it might be the most important piece. The creditor is one part of the story. The shareholder is the other part. They’re not the same.

Three institutions (BlackRock, Vanguard, and State Street) are now the largest shareholder in 88 percent of S&P 500 companies. (Fichtner, Heemskerk & Garcia-Bernardo, Business and Politics, 2017.) They hold a median stake of about 22 percent and cast somewhere between a quarter and a third of all votes at S&P 500 annual meetings. (Bebchuk and Hirst, Columbia Law Review, 2020.)

That number stops me every time I look at it. Three firms. 88 percent.

And here’s the part that matters. Bebchuk and Hirst’s research shows that these index fund managers have structural incentives to defer to corporate management. They don’t have the staff to evaluate every company independently. They don’t have the time. And their business model rewards low fees, not deep stewardship. So the largest shareholders in almost every public company in the United States are also the shareholders with the least incentive to push back on whatever management is doing.

That’s a different problem than Berzai’s creditor argument. It’s an ownership problem. The owners and the customers parted ways quietly, in the background, while everyone was watching interest rates. Cheap capital made the theater affordable. Concentrated, deferential ownership made it unopposed.

Where I’d push back on him

A couple of places. Berzai bundles ESG, DEI, and AI-everywhere together as one phenomenon. That works rhetorically but I can’t quite get the three to line up mechanistically. ESG had a clear financial driver — there was an actual yield advantage for green bond issuers, the famous greenium, that genuinely lowered the cost of capital for companies that scored well on ratings. DEI was a different animal. It was driven less by allocator demand and more by a mix of post-2020 social pressure, talent-market competition, and litigation risk after the Students for Fair Admissions ruling created new exposure. AI-everywhere was a third thing, driven mostly by valuation narrative — boards and growth-equity investors wanted to see “AI strategy” on the deck because the multiples demanded it.

All three converged on the same distortion, where companies optimized for signals aimed at someone other than the customer. But they got there by three different routes, and they’re reversing at three different speeds. ESG is unwinding through a market mechanism — the greenium has collapsed from about six basis points in 2015 to effectively zero by 2024, which is the bond market’s own quiet verdict. DEI is unwinding through legal and political pressure, with twenty-plus Fortune 500 companies rolling back programs in the last eighteen months. (Forbes, April 2025.) AI-everywhere is still in motion — the Goldman Sachs research questioning a trillion-dollar AI spend is the leading edge of what will be a longer ROI reckoning.

One other place. Berzai uses Alyssa Milano trading her Tesla as a vivid illustration of the status signal flipping. The story is real but he tucks in a phrase about people quietly replacing them with gas-powered vehicles. The registration data doesn’t support that. S&P Global Mobility tracked Tesla disposals through 2023 and found that about 70 percent of the people leaving a Tesla bought another Tesla, 13 percent bought a different EV, and only 11 percent went back to a gas car. The interesting story is that Tesla as a brand is losing position inside a category that the customer is still committed to. That’s a more useful observation than the one Berzai actually made. I’m not picking nits. I’m pointing at the texture because I think the texture matters for what comes next.

The vanishing greenium

Of all the numbers in the dossier I built for this piece, the one I keep returning to is the greenium.

The idea was that companies issuing green-labeled bonds would pay a lower interest rate than they would for an identical conventional bond. That yield gap was the financial mechanism that made the whole ESG-debt apparatus rational. If you could shave six basis points off your borrowing cost by getting an MSCI rating bump, you’d happily fund a sustainability office and publish an impact report. The math worked.

By 2024, that gap was gone in developed markets and down to about 1.2 basis points globally. (IFC-Amundi, 2025.) In some markets, green issuers now pay slightly more than conventional issuers. The bond market, which is the most clear-eyed pricing mechanism we have, because bondholders only care about being paid back — quietly stopped paying for the signal.

When the financial incentive to play a game disappears, the game ends. That’s why ESG fund flows turned negative in 2025 for the first time since Morningstar started tracking the segment because the trade stopped working.

This is the macro version of what happens inside a single company when a reset CEO arrives. Strip back. Remember who pays. Stop optimizing for things that don’t connect to a real customer or a real cash flow.

The reset is already happening. It’s happening unevenly, on different schedules, in different sectors. But the same diagnostic that worked for a single drifting company in 1993 or 2014 is now visibly working at the level of the whole market.

What this means if you’re reading this as an operator

In Part 1 I laid out a four-level framework for strategic depth. Saying it. Proving it. Being it. Owning it. Most companies live at Level 1. They have a vision statement and a budget that doesn’t match it. A few make it to Level 2 with some pilots that never scale. Far fewer make Level 3 with irreversible bets. A small handful reach Level 4 and end up owning a concept in customers’ minds so completely that they become the default.

The reason that framework still holds for this piece is that capital re-pricing doesn’t automatically take a company to Level 4. It just removes the cheap-money subsidy that let Level 1 and Level 2 companies look successful for a decade. The drift is exposed, but the reset still has to happen company by company, decision by decision.

If you run something, the test is more uncomfortable now than it was three years ago. Who are you actually optimizing for? Look at the last ten decisions your company made. Trace each one to the audience it was designed to satisfy. The customer? The board? The trade press? The institutional shareholders? The bond market? An activist nobody has met? A consultant report nobody read?

The reset CEO move only works if the answer to most of those questions is the customer. It doesn’t work if the company has spent fifteen years building muscles in the wrong places. You can’t pivot to customer-first in a quarter when the entire org chart, incentive plan, and culture has been wired for allocator-first.

That’s the part I worry about, and the part I think Berzai and Dalio both underplay. Capital re-pricing creates the incentive to refocus. It doesn’t create the capability. The companies that come through the next five years intact are going to be the ones that already had customer obsession built in, or the ones whose new leadership is willing to do what Mulally did at Ford and sell off the parts that don’t connect to the answer.

The autopilot corporation, at scale

The frame I’m settling on after writing this is simple enough. Drift isn’t a failure mode. It’s a phase. It happens to one company. It happens to a whole economy. It looks the same at both scales because the underlying mechanism is the same. When the person paying the bills stops being the loudest voice in the room, the company starts optimizing for whoever is. Given enough time and cheap enough capital, the whole economy can do this together.

The reset, when it comes, looks identical at both scales too. Strip back. Remember who pays. Ignore the noise. Make an irreversible bet on something a customer actually wants.

I started this series because I kept watching companies I respected stop being interesting. They didn’t fail in any dramatic way. They just started talking like everyone else. Slide-deck language. Category descriptions. Adjectives instead of nouns. They forgot the one-sentence answer to who is our customer and what do they pay us for.

The same thing happened to the economy. It’s been happening for a while. The interesting question, watching it unwind in real time, is which companies and which leaders are going to figure out the reset before the cycle figures it out for them.

The autopilot is off. The controls are back in human hands. What the human chooses to do with them is the part that hasn’t been decided yet.

---

Patrick Berzai’s original essay is here. Worth reading in full. Ray Dalio’s Principles for Dealing with the Changing World Order is the source of the cycle frame I borrowed from lightly. The McKinsey Global Institute and FCLT Global research on long-term company outperformance is here. The Bebchuk and Hirst work on the Big Three is in the Columbia Law Review.