COO Kumar Galhotra says machine vision inspection lacked the "nuanced judgment" to catch complex defects, costing the automaker billions, so Ford is rebuilding its human expertise layer as an active AI trainer and checker, a course correction that arrived alongside a top ranking…
Three years ago, Ford bet that AI-powered camera systems could replace much of the human judgment inside its factory quality inspections. The bet did not pay off. Over the past three years, the automaker has rehired more than 350 veteran engineers, internally known as "gray beards," Ford's nickname for its most experienced manufacturing specialists, to do the kind of nuanced defect-hunting that the machines kept missing, and to teach the machines how to do it better.
COO Kumar Galhotra put the admission on the record: "We had been relying more and more on automated quality systems and not getting the desired results. We brought back technical specialists and they hunt for failure points before a part ever reaches the plant floor" (The Independent, citing Bloomberg reporting from June 2026; Bloomberg's own write-up). The engineers split their time between leading quality reviews on the line and improving or retraining the AI systems that flagged or missed the defects, a structure that turns the human specialists into an active feedback layer for the automation rather than a replacement for it (The Independent).
The pattern the story exposes is narrow but precise. Machine-vision inspection excels at repeatable checks at high speed. It does less well when the failure mode is novel or when the defect hides in visual noise a human specialist would dismiss on instinct: a hairline crack in a casting, a tolerance drift only an experienced eye would flag, a spark on a weld the model reads as background. The Bloomberg-sourced reporting, carried by The Independent, puts the cumulative cost of that judgment gap in the billions of dollars.
That posture is now also the company line. Ford's Charles Poon has framed the operating model in plain terms: "AI is only as good as the information you use to train it" (Independent, citing the same reporting).
The story is not a wholesale retreat from automated quality. Ford's Kentucky Truck Plant still runs AI quality control alongside 3D printing on body components (Complete AI Training summary of plant reporting), and Ford has separately developed its own sensor-based quality system described by trade outlet Assembly Magazine. The reversal is targeted: where AI inspection was operating without a strong human oversight layer, the human layer is back.
The tension honest readers should hold is that Ford is also winning in places the rehiring does not reach. In 2026 the company took the top mainstream-brand spot in the J.D. Power Initial Quality Study, an annual industry benchmark that measures problems reported by new-vehicle buyers in the first 90 days of ownership, Ford's first such ranking in 16 years. That is a real recovery signal for the showroom experience. At the same time, Ford remains the most-recalled automaker in the United States, and company executives have tied that older recall volume to the same automation era that produced the inspection failures, so the picture is a company in the middle of a course-correction, not one that has completed it.
For other manufacturers weighing the same trade, the practical read is that AI quality inspection does not fail on speed, throughput, or defect-classification accuracy on well-characterized parts. It fails on the long tail of novel and ambiguous defects, which is exactly the place where human expertise compounds and where its absence is most expensive. Ford's repair is not "replace the AI with people" but "staff the AI with people whose judgment the system can learn from," and the cost of skipping that step is now on the record in the billions.
What to watch next: whether the rehired engineers' training data lifts inspection accuracy at plants beyond the early rehiring sites, whether the J.D. Power lead holds into 2027, and whether the recall volume from the automation era begins to clear. The first signals will land in Ford's next quarterly quality disclosures and in the 2027 Initial Quality Study, both of which will be a sterner test of whether human-in-the-loop is a transitional repair or a durable operating model.