The AI Layoff Regret Wave: The Graybeard Engineers Are Coming Back
"AI is a powerful tool for catching potential quality problems, but it only works in the hands of the people who use it!" In June 2026, Ford did something that surprised many: it rehired 350 senior engineers. Internally, these 350 people are called "Gray Beard" engineers. Over the past few years, Ford has aggressively pushed AI for parts defect detection and whole-vehicle quality inspection, with more than 900 AI cameras watching the production line, believing algorithms could hold the quality line. As a result, large numbers of hidden quality problems flowed into the line. Algorithms can recognize standard defects, but they cannot handle batch-to-batch material variations, latent assembly defects, or the non-standard failures that accumulate across multiple vehicle generations. Charles Poon, Ford's vice president of vehicle hardware engineering, later reflected with a plainspoken remark: "We mistakenly believed that as long as we introduced AI and absorbed the existing design requirements, we could produce high-quality products." The effect of rehiring was immediate. Ford CEO Jim Farley said the rehiring brought Ford "hundreds of millions of dollars" in cost savings. In the most recent J.D. Power initial quality survey, Ford jumped to first place among mainstream brands for the first time. The people replaced by AI were invited back.
I. Beyond Ford: Ford's Experience Is Not an Isolated Case
Ford's experience is not isolated. Last year, Australia's Commonwealth Bank planned to use AI chatbots to replace 45 customer-service staff. As a result, the AI system fell short of expectations, and the bank had to withdraw its layoff decision, apologize to the dismissed employees, and rehire them. IBM went further, directly replacing HR department functions with AI. AI can indeed handle about 94% of daily requests, but it cannot handle the remaining 6%: including ethical dilemmas and complex decisions. Nickle LaMoreaux, IBM's chief human resources officer, later said publicly that if companies do not keep investing in entry-level hiring, the talent pipeline will dry up within three to five years. IBM then announced it would triple entry-level hiring by 2026. The research firm Orgvue surveyed 1,163 executives across eight countries and regions, and found that 39% of companies had laid off staff because of AI deployment, of which 55% admitted the layoff decision was wrong. Orgvue's CEO put it bluntly: "Companies are learning the hard way that replacing people with AI without understanding the impact on employees leads to big problems." This is not a few companies misjudging — it is a widespread phenomenon.
II. What Exactly Can AI Not Replace
What exactly can Gray Beard engineers do that AI cannot learn? They have worked for decades, and a single glance at a part tells them where the hidden risk lies. This ability is not the skill of "knowing how to inspect," but an intuition — this batch of material has a slightly off variation, this assembly gap may hide a latent defect, this kind of non-standard failure appeared on the previous vehicle generation. These judgments have no standard answers, appear in no manual, and rest entirely on experience accumulated through years of mistakes, reviews, and reflection. Ford replaced them with 900 AI cameras; the cameras can recognize standard defects: wrong dimensions, color deviations, surface flaws. But they cannot recognize non-standard problems like "this batch of material feels a bit off." Because the essence of a non-standard problem is not data matching but value judgment: is this deviation worth stopping to investigate? Should the entire line be halted? Can this risk be tolerated? Behind these judgments lies a capacity for value judgment. It is not academic credentials, not a skill, not how many manuals one has read, but a capability honed by making decisions again and again in real business, bearing the consequences, and reflecting on them. Judging what is worth doing, what creates incremental value, what direction can weather the cycle. This capability is precisely the hardest to quantify, and the easiest for managers to overlook. The Gray Beard engineers Ford laid off were exactly the group with the strongest such judgment in the organization. AI can execute standardized tasks and handle 94% of daily requests. But the remaining 6% — ethical dilemmas, non-standard failures, value trade-offs — is precisely the critical part that determines an organization's level. If that 6% cannot be handled, the preceding 94% means nothing. IBM's 6% is ethical dilemmas; Ford's 6% is experiential intuition. At root they are the same thing: AI can execute, but it cannot judge.
III. The Break Between Narrative and Reality: If Ford Is a Failure from the Product Perspective
If Ford is a failure from the product perspective — AI cannot replace human judgment, and quality problems gave it the slap in the face — then Cloudflare is a failure from the capital perspective: the market no longer believes the "AI replaces people" story. In May 2026, Cloudflare released a quarterly earnings report that looked nearly perfect: revenue of $640 million, up 34% year over year; free cash flow of $840 million, a record high. But within 24 hours of the report's release, the stock plunged 24%, wiping out more than $12 billion in market value. The trigger was a decision the company announced at the same time: laying off 1,100 people, 20% of its total workforce. The CEO attributed the layoffs entirely to the success of the AI transformation — "Internal AI tool usage surged 600% over the past three months, and work that once required multiple people now needs only one person working with AI." Under that narrative, layoffs were a benefit of technological upgrading. But the market gave the exact opposite answer. Investors could not reconcile one most basic contradiction: if AI truly drove a huge efficiency gain, why did revenue grow only 34%? If the future really will be better, why lay off a fifth of the staff now? What unsettled the market even more was that gross margin fell from 75.9% to 71.2% — the first significant decline since Cloudflare went public. For the past three years, capital markets bought the AI story almost unconditionally. Layoffs could be explained as "AI replacing labor," losses as "investing in the future," and slow revenue growth as "a period of strategic investment." AI became an all-purpose basket into which any problem could be stuffed. But now the market has become picky. The logical chain of AI transformation should be: technology investment → efficiency gains → cost reduction → profit growth. But many companies reversed the chain into: announce AI transformation → cut headcount to reduce cost → maintain profit. Cloudflare showed only the last link of the chain — layoffs — without proving that the earlier links held. It did not explain which specific businesses the AI tools improved in efficiency, did not quantify the cost savings from those efficiency gains, and did not show the new revenue the AI business brought in. The deeper problem is that the efficiency gains from AI often do not turn into the company's profit, but are instead passed on to customers through a price war. When every company uses AI to cut costs, product prices fall accordingly, and the end result is that everyone uses AI, but no one's profit grows. This is the root cause of Cloudflare's gross-margin decline. And layoffs, essentially, were to offset the profit pressure from the margin decline. Only this time, the company could not tell the truth. It had to package the layoffs as a positive AI-transformation story. But the market no longer bought it.
IV. Why Companies Always Think of Layoffs First
Ford and Cloudflare: one was slapped in the face by quality problems, the other abandoned by capital markets. On the surface they are two different problems, but at the root they point to the same mistake: treating AI as a tool to replace people. Why, the moment a company adopts AI, is the first reaction always layoffs? This is actually the mental inertia left to us by the industrial age. In the industrial logic, people are cost units, and the essence of efficiency gains is doing more with fewer people. This logic held completely for the past two hundred years, because execution was indeed a scarce resource — whoever mastered more and more refined skills had the stronger competitiveness. Within this framework, AI's arrival is naturally read as a "more efficient executor," and replacing people becomes the natural choice. But in the AI age, the underlying assumption has changed. As AI can complete all standardized execution work at lower cost and higher efficiency, the value of "knowing how to use a certain tool" is falling to zero at an unprecedented speed. What is truly scarce is no longer execution but judgment: judging what is worth doing, what creates incremental value, what direction can weather the cycle. Using AI to cut headcount for cost reduction and efficiency sounds good in the short term — labor costs drop and the profit figures look better. But the company destroys the most precious thing in the organization. The people who were laid off carried in their heads years of accumulated business intuition, customer relationships, and a guide to avoiding pitfalls. These are not costs; they are the organization's memory. Ford learned this lesson at the cost of hundreds of millions of dollars; Cloudflare is learning it at the cost of $12 billion in evaporated market value.
V. The Right Direction: Humans Define, AI Executes, Humans Verify
The right direction is not AI replacing people, but AI amplifying people. After Ford rehired the Gray Beard engineers, it put them to three tasks: mentoring young engineers, participating in key design reviews, and retraining the AI tools. This is the correct division of labor: humans define the quality standard, AI executes the inspection, and humans verify the AI's results. The Gray Beard engineers were not replaced by AI; they used AI to amplify their own experience. In the past, one veteran engineer could only watch a single line; now he teaches his judgment logic to AI, lets AI complete full inspection, and himself handles only the most critical value judgments. One person's experience is amplified by AI across the entire factory. This is not a layoff; it is a capability leap. Use AI for what people cannot do, rather than replacing what people do. People shift from executor to value judge — that is the role AI should truly play. Judge what is worth doing, and let AI do it. Judge what direction can weather the cycle, and let AI execute it. Humans define the direction; AI amplifies the execution.
VI. After Technology Levels Out: As Technical Tools Grow Ever More Powerful
As technical tools grow more powerful and easier to use, "knowing how to use AI" is no longer a competitive edge. But the capacity for value judgment does not depreciate. Ford's 350 Gray Beard engineers were replaced by AI and then brought back, and now must even teach AI how to do the work. This scene itself is the best annotation: in the AI age, what is truly irreplaceable is not the person who can operate the tools, but the person with judgment. Technology becoming simpler is a good thing. It forces us to stop and think about something that busyness has long obscured: what exactly is worth doing. What was cut was not cost, but judgment. What was brought back was not manpower, but the organization's memory.
