Stagnation Slaughters. Strategy Saves. Speed Scales.
Table of Contents
- The Orthodoxy Everyone Accepted Without Testing
- Case File: Ford Hired 350 Gray Beards
- Why the Reorganization Alone Would Not Have Worked
- How Far the Other Direction Went
- The Convergence Test Nobody Ran
- The Seven Laws, Read Through Dearborn
- What This Demands of You on Monday Morning
- Frequently Asked Questions
The Orthodoxy Everyone Accepted Without Testing
By 2026 nearly every large company in America operated on the same unwritten rule: artificial intelligence absorbs coordination and judgment work, therefore experienced senior people are the expensive line item to cut. No regulation required it. No law of physics supported it. It was consensus, and consensus is the cheapest thing in the world to break. Ford broke it, and became the top mainstream brand for new vehicle quality in America.
Walk into any business and you will find two categories of constraint. The first is real: physics, metallurgy, regulatory approval, the cycle time of the press. The second is imaginary, made of beliefs that were once true, or were never true, and have gone unexamined long enough that everyone treats them as load bearing. I call the second category industry orthodoxies, and they cost more money than any line item on a P&L.
The substitution orthodoxy is the purest specimen I have seen in twenty years, because you can watch the entire market adopt it simultaneously. Manager headcount at U.S. public companies fell measurably between 2022 and 2025. Korn Ferry surveyed 15,000 professionals and found 41 percent said their organization had trimmed management layers. Gartner projected that one in five organizations would use AI to flatten structures. Meta, Amazon, Google, and Intel all moved the same direction and said so out loud.
Notice the shape of that belief. It is true for some tasks, some of the time, which is exactly the camouflage that lets an orthodoxy survive. An assumption that is flatly false dies young. The dangerous ones carry just enough supporting evidence to stop anyone from testing them properly.
Ford tested it. And here is the detail that matters most, because it is the one that gets the timeline wrong in most retellings: Ford did not test it in response to the AI spending crisis that broke into the news this summer. Ford started roughly three years ago, which puts the decision in 2023, two years before the market’s enthusiasm reached its most extreme expression. Jim Farley has described the result as an overnight success four years in the making. That is the polite way of saying the decision was made long before it could be justified in a steering committee deck.
Case File: Ford Hired 350 Gray Beards
Over three years Ford brought in 350 veteran engineers, known internally as gray beards, drawn from former employees and from suppliers, specifically to train junior staff and to reprogram AI tools that were producing unreliable output. On June 25, 2026, Ford ranked as the top mainstream brand in the JD Power U.S. Initial Quality Study, its first time in sixteen years, having climbed from 15th among mainstream brands in 2023.
Start with the starting point, briefly, because a break needs a before. In 2023 Ford sat 15th among mainstream brands in initial quality, and the company was carrying real cost from quality escapes. Recalls were running roughly $4.8 billion a year by mid 2024, according to Bloomberg. That is the whole of the before, and it is worth exactly one paragraph, because what Ford did next is the part with instructional value.
The orthodox answer was available and everyone else was taking it. Cut the expensive experienced engineers, deploy AI tooling, let the software absorb the judgment. Ford went the other way and paid for the most expensive labor in the building.
Charles Poon, Ford’s vice president of vehicle hardware engineering, gave the diagnosis to reporters in plain language: AI is only as good as the information used to train it, and Ford had not paid enough attention to the experience of the engineers who had lived through many product cycles. Read that as an engineering statement rather than a sentimental one. A model trained on incomplete institutional knowledge produces confident output that is wrong in ways only a veteran can recognize.
Ford’s chief operating officer Kumar Galhotra described what these people actually do, and the job description is the thesis of the entire case file. They hunt for failure points before a part ever reaches the plant floor. They run mandatory quality meetings. And they reprogram the AI tools to catch glitches before they occur.
That last function deserves a second read. The veterans are not competing with the AI. They are the management layer the AI never had. Every other company in the market deployed agentic tooling and assumed it would supervise itself. Ford deployed the same class of tooling and staffed the supervision.
What the Break Produced
In the 2026 JD Power Initial Quality Study, Ford recorded 152 problems per 100 vehicles, ahead of Nissan at 156 and Buick at 162. That was 41 fewer problems per 100 vehicles than the prior year, the largest year over year improvement of any mainstream brand. Ford finished third overall across the entire industry, behind only Porsche and Genesis, and second among corporations. The F-150, Mustang, and Super Duty each led their segments for the second consecutive year, and seven of ten evaluated models placed in the top three of their segments, the highest share of any automaker.
Farley’s own summary is that Ford became “the gold standard for new vehicle quality” in America. On the financial side, warranty coverage and recall costs are trending down, which Ford has characterized as a tailwind worth hundreds of millions of dollars against a company target of a billion dollars in cost savings this year.
Now the part that generalizes to your business, and it is the same lesson as the GE ultrasound case file in the pillar article. There was no new technology involved. Ford had the AI tools. It had the vision systems. The components were commercially available and the methods were proven. What changed was the assumption about whose judgment should govern them. The barrier was never engineering. The barrier was a belief about whose experience was worth paying for.
Ford broke one of the most universally held business assumptions of the decade, that artificial intelligence substitutes for expensive experienced people, by hiring 350 veteran engineers over three years specifically to train junior staff, hunt failure points before parts reach the plant floor, and reprogram AI tools producing unreliable output. Ford began this in roughly 2023, years before the market’s AI enthusiasm peaked, which makes it a genuine early commitment rather than a correction. The result was a climb from 15th among mainstream brands in the 2023 JD Power Initial Quality Study to first in 2026, its best position in sixteen years, with 152 problems per 100 vehicles, an improvement of 41 problems per 100 vehicles in a single year, and third place overall behind only Porsche and Genesis. No new technology was required. Ford already owned the AI tools and the vision systems. Only the assumption about who should operate them changed, which is the defining signature of an orthodoxy break rather than a technology advantage.
Why the Reorganization Alone Would Not Have Worked
Ford credits its quality turnaround to a multi year industrial reorganization that brought engineering, manufacturing, supply chain, and quality together earlier in vehicle development, expanded internal design reviews, deepened supplier collaboration, and strengthened software testing before code reaches vehicles. Every one of those moves was available to every competitor. The reason Ford’s reorganization produced a different outcome is that it was paired with an orthodoxy break, and a reorganization without one is just new boxes.
Anyone reading this has lived through a reorganization. Probably several. You already know the base rate. Reporting lines get redrawn, an org chart circulates, a town hall happens, and eighteen months later the business performs approximately as it did before, minus the productivity lost to the transition.
I have been on both sides of that table. Across twenty years of transformation work at Berkshire Hathaway, Illinois Tool Works, Whirlpool, and JBT Marel, I have sat in the room where the new structure gets announced, and I have been the person responsible for making one of them actually produce something. The variable that separates those two outcomes has never once been the quality of the org chart.
The clearest example I can give you is from the Refrigeration division. The analysis that turned that business was a customer-product profitability matrix showing that 74 combinations out of 1,847 were generating 140 percent of total profit, and that fifteen of them produced more than half of it. Every number in that spreadsheet had been sitting inside the company’s own systems for years. No reorganization was needed to find it. And no reorganization would have found it, because structure was never the constraint. The constraint was an unstated belief that value was spread more or less evenly across the book, which is exactly the kind of assumption nobody says out loud and everybody plans around. Once that belief broke, we deliberately shrank revenue by 30 percent and the division swung from a $175 million loss to a $48 million gain.
Notice what did the work there. Not a new system, not a new structure, not new data. A tested assumption, applied to information the company already owned. That is the same shape as Ford’s gray beards: the AI tools were already in the building, and the break was about whose judgment governed them.
Why does that keep happening to competent people? Because a reorganization changes the structure inside an unchanged assumption set. It rearranges who reports to whom without touching what anyone in the building is permitted to believe. You can merge engineering and manufacturing into one industrial system, put them in one room, and give them one leader. If everyone in that room still believes the same unexamined thing about where judgment comes from, you have built a faster mechanism for producing the same answer.
Look closely at what Ford’s new structure actually demands. Unifying engineering, manufacturing, supply chain, and quality earlier in development means catching failure points upstream, before a part reaches the plant floor. Expanded internal design reviews mean somebody in the review has to recognize the problem. Stronger software testing before code ships means somebody has to know which failures matter.
Every one of those demands the same input: pattern recognition across multiple complete product cycles. That is not a structural property. You cannot reorganize it into existence, you cannot hire it out of school, and in 2023 the entire market was actively shedding it as a cost.
This is the mechanism, and it is worth stating precisely. The reorganization created the meeting. The orthodoxy break determined whether anyone in that meeting could see what was coming. Galhotra’s description of the gray beards, hunting failure points before parts reach the plant floor and running mandatory quality reviews, is a description of the reorganization’s own success conditions being staffed.
So the honest attribution is layered rather than singular. Ford’s climb from 15th to 1st was produced by a genuine multi year industrial reorganization. But that reorganization performed fundamentally differently from the dozens of comparable reorganizations running across the Fortune 500 at the same time, and the variable that separates them is the assumption underneath. Ford’s competitors reorganized while cutting experience. Ford reorganized while buying it back at scale.
That is why this is a case file and not a news item. Structure is copyable in a quarter. A belief takes years to move, which is precisely why the imitation barrier on an orthodoxy break is an order of magnitude slower than the imitation barrier on a product feature or an org chart.
A reorganization redraws reporting lines inside an unchanged set of assumptions, which is why most reorganizations produce motion without performance change. Ford’s multi year reorganization unified engineering, manufacturing, supply chain, and quality earlier in vehicle development, expanded internal design reviews, deepened supplier collaboration, and strengthened software testing before code reaches vehicles. Those structural moves were available to every competitor in the industry, and many ran comparable programs during the same period. The variable that separated Ford was the orthodoxy break underneath the structure: while the market cut experienced engineers on the assumption that AI would absorb their judgment, Ford hired 350 veterans to supply exactly the multi cycle pattern recognition that its new upstream review structure required. The reorganization created the meeting. The orthodoxy break determined whether anyone in the meeting could see the failure coming. Structure alone is copyable within a quarter, which is why a belief change, not an org chart, is what produces a durable performance gap.
How Far the Other Direction Went
While Ford was three years into buying experience back, the rest of the market pushed the substitution assumption to a point that stopped making arithmetic sense. The clearest artifact is tokenmaxxing, the practice of treating AI token consumption as a proxy for productivity, which became an unwritten industry rule by spring 2026 and collapsed within months.
The mechanics were simple and, in hindsight, absurd. Companies needed to prove AI adoption, tokens were the only thing the vendor dashboards measured, so token consumption became the metric. Internal leaderboards appeared ranking employees by tokens burned. Heavy consumption read as high performance.
Then the bill arrived. An AI consultant told Axios in late May that a client had spent roughly half a billion dollars in a single month after failing to cap usage. Amazon shut down an internal leaderboard after employees were found pointing agents at pointless tasks to inflate their scores. Microsoft cancelled most internal Claude Code licenses as per engineer costs climbed. Uber exhausted its full year AI budget by April. On July 24, Fortune published George Sivulka’s diagnosis from the a16z newsletter, which is the best sentence written about AI deployment this year: “you just hired a million bad employees.”
Sivulka’s argument is that agents do not fail because models are weak. They fail because roughly one employee in a hundred can articulate a task clearly enough for an agent to execute it well. Everyone else produces loops, agents calling themselves repeatedly to compensate for bad instructions. The overwhelming majority of tokens purchased accomplish nothing, which is the Pareto distribution wearing a new outfit.
I include this for one reason. It is the measure of how much conviction Ford was standing against. This was not a market with mild opinions about AI and experience. This was a market so committed to the substitution assumption that it built leaderboards to rank people by how much of it they consumed, and kept the leaderboards running until the invoices forced the issue.
Ford, meanwhile, had spent that entire stretch asking a question almost nobody else was asking out loud: this cannot be the right way to run a business. That question is the whole discipline. It is unglamorous, it wins no conference applause, and in 2023 it read as a company falling behind. It was worth a sixteen year high in quality.
The Convergence Test Nobody Ran
When every major competitor behaves identically on a dimension, and neither regulation nor physics requires it, you are looking at a shared belief rather than a market requirement. That is the single fastest orthodoxy diagnostic available, it takes an afternoon, and almost nobody ran it on the substitution assumption.
Map the practices where all of your serious competitors behave the same way despite full freedom to differentiate. Regulation explains some. Physics explains some. Whatever remains is shared belief, and shared belief is raw material.
Applied to the last three years of workforce strategy, the result is uncomfortable. No regulator required companies to reduce senior engineering headcount. No physical law said a veteran engineer cannot supervise a model. The convergence was consensus, built on a contested 2025 study finding that only a small share of companies saw meaningful returns from generative AI pilots, and on vendor messaging with every incentive to encourage consumption.
Meanwhile the counter evidence was sitting in public for anyone willing to look. Nvidia’s vice president of applied deep learning has said openly that the cost of AI still exceeds the cost of human labor for many tasks. Sivulka’s framing is sharper still: humans are cheaper than tokens on average, but good tokens are cheaper at scale, and management is what converts one into the other.
Ford acted on that logic three years before it was fashionable to say it, which is the 70 Percent Rule doing exactly what it is designed to do. Ford did not have 95 percent confidence that veteran engineers would fix an AI assisted quality program. Nobody could have. Waiting for certainty would have meant starting the three year build in 2026 and collecting the JD Power result somewhere around 2029.
I have run that trade myself, on a much shorter clock. At the Refrigeration division we launched non-dispenser models in 120 days against an industry standard of eighteen months, on roughly 70 percent confidence rather than 95. Moving at 70 percent meant we would know within 120 days whether we were wrong. We were not, and the decision was worth $8 million in first year profit. The study that would have lifted us to 95 percent confidence would have cost $12 million in profit we never would have booked. Ford ran the same calculation on a three year horizon instead of a 120 day one, against a market consensus pointing the opposite direction the entire time, which takes considerably more nerve than my version did.
The Seven Laws, Read Through Dearborn
The seven laws of orthodoxy smashing were derived from case files spanning decades. Ford confirms five of them cleanly, and the AI spending correction confirms the sixth and seventh.
Law One, Hidden Opportunity. The biggest opportunities sit behind the most deeply held beliefs. The most universally held business belief of the decade was that AI substitutes for expensive experience. The largest available return went to the company willing to test it.
Law Two, Customer Truth. Look for compensating behavior. Employees gaming token leaderboards, and then hoarding institutional context as a job security tactic, are workarounds. Every workaround is a signed confession that an orthodoxy is costing somebody something.
Law Three, Organizational Resistance. Resistance rises with the success the orthodoxy produced. Notice how difficult it was in 2023 for any executive to say publicly that they were hiring senior engineers back. It read as an admission rather than a strategy, which is identity lock behaving exactly as described.
Law Four, Market Timing. Too early kills as reliably as too late, and Ford moved early. What made the timing survivable is that receptivity was already built internally: the quality gap gave the organization a reason to accept an unorthodox answer, and the company had the patience to let a three year build mature before the scoreboard arrived.
Law Five, Competitive Response. Denial, dismissal, analysis, then bad copying. We are in the analysis stage right now. Expect a wave of poorly executed veteran hiring programs within twelve months, run by companies that copy the form without understanding the mechanism, and that will staff the headcount without building the upstream review structure that gives the headcount somewhere to act. That gap is the runway, and it is open today.
Law Six, Cascading Impact. Break one and the attached ones destabilize. Ford’s break on experience made adjacent assumptions visible and breakable in turn: that suppliers are vendors rather than a talent pool, that quality is a manufacturing problem rather than a design problem, and that AI tools ship finished rather than requiring continuous human retuning.
Law Seven, New Orthodoxies. Today’s break becomes tomorrow’s orthodoxy, and the warning is aimed at everyone about to imitate this. Today’s innovation is tomorrow’s prison. Within thirty six months, hire the veterans back will itself be an unexamined rule, defended with the additional authority of having worked. Put the audit on the calendar now, and put your own recent wins inside its scope.
What This Demands of You on Monday Morning
Five moves, all available this week, none requiring a consultant or a six month study.
1. Run the convergence test. List the practices where every serious competitor behaves identically. Strike the ones explained by regulation or physics. What remains is shared belief, and it is your inventory.
2. Grade the evidence behind your top five metrics. Take the five numbers your leadership team reviews most often and grade each one strong, moderate, weak, or none, based on actual evidence that it predicts business outcomes. If a metric survives only because a vendor supplies the dashboard, you have found a live orthodoxy.
3. Find your gray beards before you cut them. Identify the people who have lived through multiple complete cycles of your product, process, or market. Ask them what the automation is getting wrong. They already know. Nobody has asked, because asking contradicts the story the transformation program is telling.
4. Audit your last reorganization for a missing break. Pull up the structure you implemented most recently and name the assumption it tested. If you cannot name one, you moved boxes. That is recoverable, but only if you go find the belief the new structure actually requires and check whether anybody holds it.
5. Put an expiration date on your own best idea. Whatever break you executed most recently, schedule the audit that will attack it. The early warning to watch for is people building workarounds around your innovation. When that starts, the break has calcified and the framework now applies to you.
Ford’s competitors had access to the same AI tools, the same consultants, the same case studies, and the same freedom to reorganize. Several of them reorganized during the identical window. One of them stopped to ask whether the assumption underneath the industry’s favorite move was actually true, and that company is now the gold standard for new vehicle quality in America.
Your industry has a belief in it right now that everyone accepts and nobody has tested. Somebody is going to test it. The only question is whether they work for you or against you.
Frequently Asked Questions
Why did Ford hire 350 gray beard engineers?
Ford hired 350 veteran engineers over roughly three years, drawn from former employees and from suppliers, to train junior staff, hunt failure points before parts reach the plant floor, and reprogram AI tools that were producing unreliable output. The company had concluded that its AI systems were only as good as the experience informing them. Ford went on to rank as the top mainstream brand in the 2026 JD Power U.S. Initial Quality Study, climbing from 15th among mainstream brands in 2023 to first in 2026.
What is orthodoxy smashing?
Orthodoxy smashing is the systematic identification and testing of the unwritten rules an industry accepts without evidence. It differs from continuous improvement because it attacks the assumption behind a process rather than the process itself. The method runs four steps: identify 15 to 25 unwritten rules, challenge them by grading evidence against impact, create options on new dimensions, and validate through bounded pilots with criteria written before launch.
Why do most reorganizations fail to change performance?
Because a reorganization redraws reporting lines inside an unchanged assumption set. New structure plus old beliefs produces a faster route to the same answer. Structural changes are also copyable within a quarter, so they rarely create durable separation. A reorganization produces a fundamentally different outcome only when it is paired with a tested change in the underlying assumption, which is the pattern visible in Ford’s quality turnaround.
What is tokenmaxxing?
Tokenmaxxing is the practice of treating AI token consumption as a proxy for employee productivity, often enforced through internal leaderboards ranking staff by tokens used. The term entered wide circulation in April 2026. Critics argued that employees would maximize any metric management tracked, producing higher costs and lower quality output rather than genuine productivity gains, and the practice collapsed within months as costs surfaced.
How do you tell an industry orthodoxy from a real constraint?
Real constraints are physics, metallurgy, regulation, and material limits, and they can be named specifically. Orthodoxies cannot. Require anyone rejecting an outside example to finish the sentence: that would not work here because the mechanism that makes it work is X, and X is absent in our context. Nine times out of ten they cannot name the mechanism, because they rejected the surface form rather than the principle.
About the Stagnation Assassin
Todd Hagopian is a Fortune 500 transformation executive who has generated $3B+ in shareholder value across Berkshire Hathaway, Illinois Tool Works, Whirlpool, and JBT Marel, where he serves as VP of Global Product Strategy. Known as The Stagnation Assassin, he is the author of two published books: The Unfair Advantage: Weaponizing the Hypomanic Toolbox and Stagnation Assassin: The Anti-Consultant Manifesto. His blog is published in 15+ languages and read by operators worldwide. Bring him to your stage via his speaking page or connect with him on LinkedIn.
Is your reorganization missing its orthodoxy break? Most structural change programs move reporting lines inside an assumption set nobody has tested, which is why they generate motion without separation. An Orthodoxy Smashing diagnostic runs the identification and challenge steps against your business: the convergence audit, the evidence grading, and the shortlist of five to seven assumptions worth attacking first, with the cost of carrying each one attached. Book a confidential Orthodoxy Smashing diagnostic with Todd and find out which one you are paying for.