Ford rehired 350 engineers after AI vehicle quality failed, admits VP Charles Poon

Ford has admitted it had to rehire 350 experienced engineers after its AI systems delivered the wrong vehicle quality. Charles Poon, Ford’s VP of vehicle hardware engineering, told reporters that the automaker mistakenly believed it could “swap in AI” and still produce a high-quality product. The admission, first reported by The Verge, is a blunt reminder that AI in manufacturing is not just a software upgrade, it is an end-to-end change to how products are designed, verified, and approved.

That “350” number matters, because rehiring experienced engineers is not a small operational tweak. It is a signal that something more fundamental than a minor model inaccuracy went wrong. When an automaker has to pull in seasoned hardware engineers again, it usually means the gap between the expected output and actual quality is too big to bridge with quick fixes. Poon’s framing also points to the core failure mode: Ford did not treat AI as an additional tool with guardrails. Instead, it appears to have assumed the AI could replace or compress steps in the quality process without degrading outcomes.

To understand why this is such a big deal, you have to zoom out to how car quality works in the real world. Vehicle hardware engineering is full of interlocking tolerances and validation steps. The “quality the company expected” is not one metric. It is the aggregate of performance, reliability, and manufacturability results that ultimately have to satisfy safety, compliance, and customer expectations. In that environment, AI systems can help with inspection, prediction, optimization, or automation. But if the AI becomes a de facto decision-maker or substitutes for validation work, it changes risk. And in a product where recall and regulatory scrutiny are existential, even small process errors can become expensive quickly.

This is also where incentives bite. Automakers operate under relentless pressure to reduce costs, speed development cycles, and scale production while managing complex supply chains. AI is attractive because it promises acceleration: faster identification of defects, more efficient testing, and fewer manual checks. But the admission from Poon suggests Ford’s internal assumption was that AI could “swap in” without sacrificing a baseline of quality. That kind of leap can look good on a roadmap. It can look even better in board updates. Until it collides with reality, and reality forces a rehire.

The regulatory background matters even if the report excerpt does not specify a regulator by name. Vehicle manufacturing exists in a compliance universe: manufacturers need to prove that vehicles meet standards and that processes are reliable. When quality misses expectations, companies do not just fix things to satisfy customers. They also need to preserve trust with regulators, auditors, and insurers, and they need to show a defensible quality chain. Even when AI is used for improvements, regulators will care whether the process produces consistent results and whether failures can be detected and contained. In other words, “AI produced the wrong quality” is not only a production issue. It is a governance and assurance issue.

Now consider the operational second-order implications. Rehiring 350 experienced engineers likely means Ford spent time and money reversing course, rebuilding expertise, and re-centering the quality process around proven engineering practices. That does not just impact the specific AI deployment. It can also slow downstream programs, because engineering teams are finite. If you pull experts back into the loop after an AI-driven workflow fails, you are also pausing where those experts could have been working on new vehicles, new components, or new manufacturing improvements.

And there is a board-level implication for every executive reading this. This kind of admission can affect how AI risk is discussed internally. Boards that once treated AI as a productivity lever may shift to treating it as a risk category that needs measurable controls, auditability, and performance thresholds. The headline from The Next Web, built on Poon’s comment reported by The Verge, effectively tells leadership teams: AI can fail in production quality, and when it does, the cost is not theoretical. It is people, time, and rework.

For other operators and investors, the takeaway is clear: safety-critical manufacturing is unforgiving. AI can be part of the system, but it cannot casually replace quality assurance and validation. Ford’s rehiring of 350 engineers after AI “got vehicle quality wrong” is a concrete case study in what happens when a company believes it can swap in new technology while keeping output quality unchanged. The strategic stakes go beyond Ford. Any company trying to industrialize AI in hardware, regulated products, or high-stakes processes now has a real reference point for how quickly ambition turns into expensive remediation.