What it was for
The front half of Design Thinking — sitting with users, surfacing the unstated, converging on a problem worth solving — existed because building the wrong thing was the most expensive mistake available, and humans are reliably wrong about what other humans need until they look.
The verdict
ENDURES — and appreciates. Framing the problem is pure outer-loop work: judgment about people, applied before a single revolution of the engine spins. Cheap construction makes it more valuable, not less, because the engine will now build the wrong thing faster and more convincingly than any team in history. The quality of understanding is the quality ceiling of everything downstream.
What changes
Not the skill — the position. Framing loses its monopoly on a gated "phase one" and becomes a continuous posture, refreshed every time evidence flows back through the outer loop. One trap is worth naming in advance: synthetic users. Machine-simulated research is seductive precisely because it is frictionless, and it fails exactly where this method earns its keep — at the unstated, the embodied, the thing nobody knew to say.
The strongest objection
AI can interview, transcribe, and synthesise research at scale — surely empathy is automatable too. The tools, yes; conceded gladly. The judgment about which human signal is real and which is performance remains the practitioner's, and there is no telemetry for it.
Corroboration — the specification becomes the product (IBM, 2026)
Kim Bartkowski, writing from IBM, reaches this verdict from the economics rather than the failure mode. This entry argues framing appreciates because the engine will build the wrong thing faster and more convincingly; she argues it appreciates because falling build costs move competitive advantage from production to intent. The question that matters stops being how to build something and becomes what to build, for whom, and why.
Two things are worth keeping. She names the mechanism this card leaves implicit: when every organisation has the same models and the same tools, the remaining differentiator is context, behavioural insight, and trust, since the generation itself is available to everyone. And she extends the position past framing into defining how intelligent systems behave: how a model explains its reasoning, how uncertainty is communicated, when automation should be used at all.
IBM Design formalised Enterprise Design Thinking, the practice this card audits, which makes the convergence worth noting. (Bartkowski, "When Software Is Generated, Human Understanding Becomes the Differentiator," Bootcamp, July 2026.)