Ant Murphy opens with an eyebrow-raising McKinsey number:

McKinsey reports that 88% of organisations say they “use AI” but only about 1% have mature AI deployments delivering real value.

Murphy’s explanation for the gap is familiar: the diffusion of innovation, Geoffrey Moore’s chasm between early adopters and the majority, now applied to AI. What’s less common in the AI discourse is a behavioral explanation for why the adoption keeps stalling. Murphy:

AI is personal. It’s not another tool, to some it’s viewed as a replacement. “AI attacks our identity in a way that most software doesn’t” — Vikram Sreekanti

That resistance shows up in the record: a friend’s “I didn’t sign up for this”. Claire Vo described designers as the most resistant to change in the EPD triad, vocal AI opponents with little appetite for campaigning for resources. None of it is irrational. Daniel Kahneman and Amos Tversky found that humans weigh losses about twice as heavily as equivalent gains. Years of accumulated craft become our identity. AI doesn’t ask you to learn new tools; it asks you to renegotiate what made you worth hiring in the first place. The reskilling conversation treats that as a capability problem. Identity problems don’t resolve themselves through training on new tools.

Murphy on what that requires:

Surviving a paradigm shift like this is less about what your product does […] Instead it’s about you adapting to the change.

The 88% are held back by what AI is asking them to let go of. Murphy’s argument is that organizations clearing the chasm are doing the internal work first—on process, on how teams function—before it shows up in the product.

There’s an old relationship adage that you can’t be a good partner to someone until you’ve worked out your own stuff first. I think Murphy’s argument is the organizational equivalent.