Why Ford Rehired

Ford has added more than 350 experienced engineers over the past three years after its automated quality systems failed to deliver the results the company expected. Inside Ford they are called gray beards. Some are former Ford employees. Others came from suppliers.

Ford says it added the specialists “through internal promotions or new talent” to work alongside newer team members. Headlines have called it rehiring people AI replaced but Ford has not used that word, and the reporting does not establish that these specific roles were cut.

Charles Poon, Ford’s vice president of vehicle hardware engineering, told reporters that AI is a fantastic tool and only as good as the information used to train it. He was more direct about the error: “Mistakenly, we thought that by just introducing artificial intelligence and ingesting the design requirements that we had, that would produce a high-quality product.”

Poon also said Ford had not paid enough attention in prior years to the experience of its most knowledgeable engineers, the ones who had been through many product cycles.

Kumar Galhotra, Ford’s chief operating officer, said the company had been relying more and more on automated quality systems before recognizing the approach was not working.

Ford then ranked first among mainstream brands in the 2026 J.D. Power U.S. Initial Quality Study, its first time since 2010.

Ford’s explanation

Ford describes a training data problem. Poon’s version: enhancing the automation and machine learning tools required making sure they were trained by the most experienced individuals.

Part of that holds up plainly. The specialists do reprogram the AI tools that fell short.

The jobs

A missing corpus has a fix with an end date. Sit the veterans down, extract the failure modes they carry, feed the models, thank them.

But Ford built something with no end date. The specialists run mandatory meetings on quality concerns and hunt for failure points before a part reaches the plant floor. They conduct regular design reviews to identify problems before vehicles reach production. They train junior staff. Galhotra put the shift as moving from a find-and-fix mentality to preventing issues before they occur.

A standing design review is a verification loop the organization has decided to keep running.

Where the corpus story runs out

Tacit knowledge can be captured up to a point. What a senior engineer knows about how a joint fails under a particular thermal cycle can be written down, and should be.

Judgment applied to an unanticipated case cannot. A reviewer looks at a novel configuration and says it will not hold, for reasons that emerge from the thing in front of them. Enumerate those cases ahead of time and you would have automated the review already.

The corpus framing implies a completion state, where enough capture makes the humans optional. Ford’s remedy points elsewhere. Mandatory and recurring is what you build once you have concluded the checking does not stop.

Automating a quality inspection function means automating a verifier, which removes the capacity to tell whether the automation works.

Ford’s specialists hold the ability to tell the machine it is wrong.

The order

Poon’s admission about prior years is the sharpest thing either executive said. Ford’s assumption was reasonable and its sequence was wrong. Capture the expertise, then automate, and the program is defensible. Automate on the assumption that design requirements are sufficient, and you spend three years buying judgment back from suppliers and internal promotions.

Each step is the precondition for the next. Skipping one relocates its cost to a later point, larger, with fewer options available. Ford turned a knowledge problem into a three-year staffing program, and that program ran alongside more than $1 billion in expected warranty and material costs this year and a quality reputation to repair.

Mentorship

Ford is explicit that mentorship is part of the assignment. The specialists work alongside newer team members and train junior staff who never absorbed the institutional knowledge.

This is how senior judgment gets built. Less experienced people work real problems while someone holding the judgment watches and corrects. The routine cases are the training ground, and automation takes them first.

An organization that automates routine work and then loses its seniors breaks the pipeline at both ends. Nobody holds the judgment and nobody acquires it. Ford is paying to rebuild both.

Headcount models calculate savings against the cost of the people. The judgment training pipeline appears nowhere in the model.

Ford is still deploying AI

On an autumn 2025 earnings call, Galhotra said Ford was systemically deploying AI across the entire industrial system, including 900 AI-powered cameras across its plants to detect quality issues at the source. Ford kept the cameras. Jim Farley told Bloomberg TV that Ford has AI tools for vision systems, and that most of it comes down to team members paying attention to small details.

Ford topped the mainstream J.D. Power rankings with the automated systems still running. Experienced people now sit between the systems and the product.

Experienced engineers sat on Ford’s books as execution capacity, a cost line. Their function was judgment, which is what makes execution capacity safe to deploy.

Ford is not unusual in getting the order wrong. CNBC reported Robert Half data showing 32 percent of U.S. hiring managers eliminated a role primarily because of AI and later rehired for the same or a similar position. Robert Half’s own summary puts it as more than 3 in 10, and notes the two most common reasons given: the role required institutional knowledge or context AI could not replace, and it involved relationship management AI could not replicate.

The arguments here about judgment as the layer that cannot be purchased, and about sequence in organizational automation, run through my book, Coherence, arriving this Fall. If you want to follow the thinking as it develops, join the list at coherise.com. The one-page decision tool from the book is the first thing I send.