Ford hired, promoted, or brought back more than 350 experienced engineers to fix errors introduced by its AI-driven design and production systems, executives said this week, crediting the rebuild for the company's first No. 1 finish in JD Power's initial quality ranking in 16 years. The admission is unusually direct for a Detroit automaker: Ford's automated systems weren't as reliable as it had assumed, and reversing the slide required pulling veteran humans back into the loop.
The story Ford is telling cuts against the prevailing AI-in-manufacturing pitch. Charles Poon, Ford's VP of vehicle hardware engineering, said the company believed that plugging AI into existing design requirements would produce high-quality vehicles on its own. It didn't. Some of Ford's most experienced personnel had already left, taking institutional knowledge with them before it could be encoded into the automated tools meant to replace their judgment.
Poon said the rehired and promoted engineers, numbering more than 350, are now mentoring younger staff and rebuilding the data pipelines that feed Ford's AI training. The effectiveness of those models, in Ford's telling, depends almost entirely on the quality of the underlying data, and the data wasn't capturing what veteran engineers knew from working through multiple vehicle-development cycles.
“Mistakenly, we thought that by just introducing artificial intelligence and adjusting the design requirements that we had, that that would produce a high-quality product.”— Charles Poon, Ford VP of vehicle hardware engineering
Key facts
- 01Ford hired, promoted, or brought back over 350 experienced engineers to correct errors made by its automated design systems.
- 02The automaker was named No. 1 in JD Power's initial quality ranking for the first time in 16 years.
- 03Ford created a dedicated 40-person software quality assurance team focused on preventing defects before they occur.
- 04The company added more than 100,000 new AI-powered automated tests to revalidate software changes late in development.
- 05Ford VP Charles Poon said the company underestimated the institutional knowledge of veteran engineers when it leaned on AI.
The context matters. Ford currently leads the US auto industry in recalls, and its quality scores have slipped for several years running. Botched launches of the Explorer and Aviator, supply-chain shocks from the covid pandemic, and a growing backlog of recalls compounded the problem. The JD Power top spot — the first since 2010 — is the first concrete external signal that the rebuild is working.
Ford COO Kumar Galhotra framed the underlying failure as structural rather than purely technical. Different departments operated in silos. The company defaulted to a "find and fix" approach: catch the defect after it shipped into the pipeline, then patch it. That worked for individual issues but did nothing to stop the next one.
Galhotra said Ford is now pushing hardware engineering, software, manufacturing, and supply-chain teams to work in tighter coordination, with metrics focused on leading indicators rather than after-the-fact defect counts. The cultural shift, he argued, is as important as the staffing one.
On the software side, Ford built a dedicated 40-person quality assurance team with one job: prevent software defects before they reach customers. Poon said the company had been catching software bugs too late in development because it wasn't iterating quickly enough early on. But Ford also can't borrow the consumer-electronics habit of shipping broken software and patching later — vehicles are safety-critical from the moment they leave the lot.
To bridge that gap, Ford expanded its automated testing capabilities with more than 100,000 new AI-powered tests designed to stress software under edge-case conditions. The pitch is that with enough automation, even a late code change can be re-validated end-to-end before delivery, instead of becoming a recall six months later.
The skeptic's read is that 100,000 automated tests and a 40-person QA team are still bets on more automation to fix problems that more automation helped create. Ford hasn't disclosed how many of the AI-driven design errors made it into shipped vehicles, how many recalls trace to the period when veteran engineers were leaving, or what the unit economics of the rebuild look like. JD Power's initial quality survey measures owner-reported problems in the first 90 days — a useful signal, but not a long-term durability verdict.
Ford's experience is a data point worth more than a press release for every executive currently being told that AI will compress engineering headcount. The lesson here isn't that AI doesn't work in heavy industry — Ford is doubling down on automated testing, not retreating from it. It's that automated systems inherit the gaps in their training data, and when the people carrying the tacit knowledge walk out the door before that knowledge is captured, the AI inherits nothing useful. The companies that win the next decade of industrial AI will be the ones that figure out how to transfer human expertise into models before they thin out the humans, not after.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
