AI In Production: Where It Works, Where It Doesn't | Week 13 to 19 Jul 26

This week shows why AI adoption isn't one-size-fits-all. Two very different outcomes—one from Netflix, one from Ford—illustrate a critical gap between where generative AI excels and where it still needs human judgment to avoid costly mistakes.

The win this week

Netflix deployed generative AI across approximately 300 titles to handle a specific production bottleneck: creating complex visual sequences that take time and money to produce traditionally. We're talking about crowd scenes, historical battle sequences, and establishing shots—the kind of work that's visually intensive but follows predictable patterns.

The key to Netflix's approach was integration, not replacement. AI generated these sequences during post-production, but human teams reviewed and approved the output before it reached viewers. Netflix kept the repetitive visual work on machines while preserving human oversight at the approval stage. The result was faster production cycles and lower costs without sacrificing the quality bar.

This matters because Netflix applied AI to a well-defined task with measurable inputs and outputs. The company knew exactly what it needed (crowds, battle scenes, establishing shots), had clear quality standards, and built a workflow where AI handled the grunt work while humans stayed in control of what shipped.

(Source: The Verge, 16 Jul 2026)

The lesson this week

Ford took a different path and encountered a different outcome. The company attempted to automate quality testing using AI, replacing experienced engineers who had deep expertise in catching manufacturing and design flaws. The AI quality testing system failed to deliver what the company needed.

Ford's response was to rehire those engineers, signaling that the AI system couldn't do the job alone. Quality assurance in automotive manufacturing isn't just pattern recognition—it requires judgment calls about risk, precedent, and edge cases that a new system hadn't been trained to handle. Ford discovered that removing human judgment from a high-stakes, consequence-heavy role created a gap that AI couldn't fill.

(Source: Ratopati, 30 Jun 2026)

What to do about it

Netflix's success reveals why the AI win worked: the company applied generative AI to a discrete, repetitive production task where the output could be inspected and approved before release. There was a clear feedback loop, measurable success criteria, and a human checkpoint. AI became a force multiplier for defined work, not a replacement for decision-making.

Ford's failure teaches a harder lesson about what AI isn't ready for yet. Quality testing—especially in industries where failures have safety and legal consequences—requires experience, context, and judgment. It's not enough to have a system that recognizes patterns; you need one that understands when patterns don't apply. Removing experienced engineers before the AI system had proven itself created a costly gap.

The pattern across both stories is this: AI works best when it augments skilled people doing well-defined tasks, not when it replaces judgment-heavy roles without proof of capability first. If the work requires specialists to say "no" or "that's different," move slowly. If the work is about generating variations of something humans can easily check, scale it up.

Disclaimer

This article is based on LinkFeed Issue 35 (13 Jul to 19 Jul 2026) — two verified case studies (one AI win, one AI failure). Sources: AI Win: The Verge, 16 Jul 2026 · Lesson: Ratopati, 30 Jun 2026. For informational purposes only; verify critical claims at the source.

Subscribe to LinkFeed weekly intelligence at linkifico.com/linkfeed

Need strategic AI guidance for your business? Book a Linkifico Assessment at linkifico.com/contact