The Seat Bracket That Couldn't Ship

I've been thinking about a seat bracket.

Not just any seat bracket. One that looks like it was grown rather than designed. Picture an aluminum lattice, organic and airy, the kind of structure you'd expect to find in bone or coral. Forty percent lighter than the original. Twenty percent stronger. When GM's engineers first saw what Autodesk's generative AI had produced, I imagine there was a moment of genuine wonder. Here was a part that had evolved past human imagination.

Then someone asked the obvious question: How do we make this?

The part doesn't belong here. That's the whole problem.

GM's supply chain has been stamping steel for decades. Supplier relationships, tooling investments, production lines. All optimized for a particular way of making things. The bracket required 3D printing, a technology GM uses but hasn't scaled for mass production. The part was brilliant. The system that would need to produce it existed in a different reality.

The bracket appears to have remained a proof-of-concept. GM mentioned motorsports applications first, consumer vehicles later. As of now, no production plans have been announced.

The Pattern Worth Noticing

This got me thinking about a pattern I keep seeing. Forty-two percent of companies abandoned most of their AI initiatives in 2025.¹ Not because the AI failed. Usually it worked fine. The constraint was absorption. The distance between what AI can imagine and what organizations are built to receive.

The meeting ended. The follow-up never came.

Here's what makes it interesting. Around the same time, Apple was experimenting with metalenses. Ultra-thin optical components designed through machine learning. Different industry, similar ambition. But Apple controls its supply chain, its manufacturing, its product integration. When their AI imagined something new, they could actually make it. Analyst projections have metalenses in Face ID sensors by 2026.

Same technology wave. Different organizational capacity to ride it.

What We Keep Missing

We keep asking whether AI is ready. Whether it's smart enough, capable enough. The seat bracket suggests we're asking the wrong question. GM's AI was plenty smart. It designed something genuinely better.

The problem was that GM the organization couldn't metabolize what GM the AI lab produced.

This is the gap that doesn't show up in demos or vendor pitches. It's not a technology gap. It's a muscle memory gap. An operating model gap. The kind of gap that takes years to close, because it's not about upgrading software. It's about rewiring how an organization makes decisions, builds relationships, moves resources.

The bracket sits somewhere between triumph and cautionary tale. A reminder that before asking what AI can do for us, we might ask what we're actually built to receive.

Forty percent lighter. Twenty percent stronger. Still waiting for an organization ready to make it real.

---

References

Footnotes

1. S&P Global Market Intelligence. (2025). "AI experiences rapid adoption, but with mixed outcomes." CIO Dive ↩