Two stories this morning sit in perfect tension: one carmaker discovering the limits of automating without people, another pushing humanoid robots deeper onto its factory floor. Together they're a sharp snapshot of where "AI on the production line" actually stands — useful, advancing, and still very much in need of humans who know what they're doing.

Ford's lesson: AI is only as good as the people training it

The headline doing the rounds — "Ford's AI failed" — overstates it, but the real story is more interesting. According to Bloomberg, Ford has spent the last three years quietly rehiring and promoting around 350 veteran engineers — the "gray beards" with decades of hard-won know-how — after discovering that AI and automation alone weren't delivering the quality it expected.

What those veterans actually do is the point. They mentor junior staff, lead mandatory design reviews, catch failure points before a part ever reaches the floor — and, crucially, help train and supervise the AI quality tools themselves. Ford's VP of vehicle hardware engineering, Charles Poon, put it plainly: "Artificial intelligence is a fantastic tool, but it's only as good as the information you use to train it." The company admitted it "didn't pay as much attention as we should have to the experience of our most knowledgeable engineers." The result of bringing them back: Ford topped J.D. Power's latest quality ranking among mainstream brands.

So this isn't "AI doesn't work." It's the garbage-in, garbage-out truth made concrete: automation deployed on top of thin or fading human expertise produces thin results. The veterans weren't rehired to replace the AI — they were rehired to make it good. It's the most grounded entry yet in the AI-and-jobs story we've followed all week, the counterpoint to the cut-staff-and-automate narrative: sometimes the expertise you let go is exactly what the AI needed.

BMW's lesson: but the robots really are coming

For the other side of the coin, look to South Carolina. BMW has announced it's deploying Figure's new "Figure 03" humanoid robots into logistics work at its Spartanburg plant — picking unsorted parts from containers, sorting them into sequencing trolleys, and prepping them "just in sequence" for the assembly line. It follows an earlier pilot in which the previous-generation Figure 02 robots reportedly supported the build of more than 30,000 vehicles over about ten months.

Worth keeping the salt handy: this is a joint announcement from BMW and Figure, so the claims about the new robots' abilities are the companies' own, the robot count isn't disclosed, and a press release is a sales document as much as a status report. But the direction is real and not hype: humanoids are moving from demo videos into genuine, if narrow, factory roles. Set beside Ford, it's the honest picture — automation is advancing fast and hitting real limits, often inside the same industry, at the same time.

Also today

Apple may skip a chip generation for AI. Bloomberg's Mark Gurman reports that Apple plans to skip its high-end M6 Mac chips and jump to an "AI-focused" M7 line, pulling the generation forward to prioritise faster on-device AI. Treat it as a well-sourced report, not an Apple announcement — but it fits a clear trend: the race to run capable AI locally, on your own device, rather than in the cloud.

And the Anthropic–Alibaba row stays one-sided. A day after Anthropic accused operators tied to Alibaba's Qwen lab of "illicitly" distilling Claude (an allegation, still unproven, that we covered in depth), Alibaba has still issued no official response. The only pushback so far is commentary — a Chinese industry figure dismissing it as "technological hegemony anxiety" and a foreign-ministry line calling such claims "groundless" — not a company denial. The silence itself is now part of the story.

The throughline for a Friday: the most honest place to stand on AI and work isn't "it replaces everyone" or "it changes nothing" — it's Ford and BMW in the same breath. The robots are real, and so is the expertise they still can't do without.