A Ford VP publicly called it a mistake to think AI alone could catch quality problems. The 350 experienced engineers brought back to fix the fallout sit awkwardly beside the CEO's claim that AI will replace half of US white collar work.
Ford's vehicle hardware engineering chief made an admission this week that executives rarely volunteer in public: the company had "mistakenly" believed that introducing AI and tightening design requirements would, on their own, produce high-quality vehicles. It was wrong. The fix involved bringing back the experienced engineers the rollout had sidelined.
Charles Poon, Ford's VP of vehicle hardware engineering, told reporters at The Verge that the assumption failed because AI systems don't know what they haven't been trained on. The underlying reporting, summarized by Futurism, describes the consequence: Ford has since rehired, newly hired, or promoted roughly 350 experienced engineers to handle quality problems the AI failed to catch.
That contradiction is the story. Ford has simultaneously topped JD Power's initial quality study for the first time in nearly two decades, and still leads the US in recalls this year, slipping in the same firm's dependability rankings. CEO Jim Farley has been telling industry audiences that AI will "replace literally half of all white-collar workers in the US," per Business Insider's reporting. Yet the company's response to a high-profile AI quality failure has been to put experienced humans back in the loop.
The mechanism here is not generic "AI doesn't work." It is narrower and more useful: in safety-critical production, the people who have seen thousands of edge cases know what AI systems haven't been trained on. When those engineers leave before their institutional knowledge is transferred to the systems meant to replace them, the AI inherits a blind spot it cannot recognize, and quality problems compound faster than the AI can catch them. Ford's tacit-knowledge gap is visible in the numbers: roughly 350 rehired, newly hired, or promoted engineers, against a workforce that is down about 5,000 versus 2020.
Ford isn't retreating from AI. It's adding to it. The company has layered in more than 100,000 new AI-powered tests to identify edge cases and stress software systems, per The Verge's reporting. The pattern is concrete: an AI rollout that assumed human judgment was replaceable, followed by a remediation that pairs more AI with the human judgment the rollout had assumed away. As Inc. framed the episode, the miscalculation was not investing in AI. It was treating the experienced engineers around it as optional.
What to watch next is whether the 350-engineer rebuild sticks, and whether JD Power's dependability rankings move in the same direction as its initial-quality top spot. If they don't, the contradiction between Farley's rhetoric and Poon's admission becomes harder to wave off as a transitional phase.