Adoption: Where the Value Actually Lives
A model that nobody uses is worth zero. McKinsey is clear that adoption and scaling — not the build — is where most transformation value is won or lost.
A model nobody uses is worth exactly zero. Not less than you hoped — zero. The value was never in the building; it was always in the using.
This is capability six, the one everyone agrees with and nobody funds. The build feels like the achievement: the model trained, the tool shipped, the launch announced. But value isn't created when something is built — it's created when people change how they work and keep working that way. McKinsey is emphatic that adoption and scaling is where most of the value actually lands, and that it routinely gets a fraction of the attention the build does. The arithmetic is unforgiving: a tool that would save every user an hour a day, used by nobody, saves nothing; used by ten percent, it captures ten percent of its value. The gap between what you built and what gets adopted is usually where the majority of the promised return quietly disappears.
Pilot purgatory is the modern shape of this
The gen-AI years gave this leak a name: pilot purgatory. Something like two-thirds of AI efforts get stuck between a promising pilot and real scaled use — and the blockers are almost never the model. They're operating-model inertia, workflows too rigid to absorb the change, and the absence of any measurement that would prove value and justify the push. Scaling, in other words, is its own discipline, distinct from and harder than the build — a climb up a ladder most programmes never finish:
Building is visible and adopting is patient. A launch is a date you can put on a slide; adoption is months of change management, retraining, redesigning incentives, and fixing the friction that keeps people on the old way. It doesn't photograph well, so budgets pile onto the build and trail off right when the harder, value-bearing work begins. The company that treats adoption as the main event — resourcing it like one — is the one that captures the prize the roadmap promised.
Where it goes wrong
Declaring victory at launch and counting the build as the win. The tool exists, the box is ticked, the team moves on — and the usage chart flatlines at fifteen percent while everyone celebrates. Measure adoption, not delivery. If your dashboard tracks what you launched but not what's actually used, you're measuring the cheap half and ignoring the expensive one.
Try this
Take something your organisation built and rolled out, and find the real number: of the people meant to use it, how many actually do, daily? Put that next to what it cost to build. The gap between the spend and the usage is the value adoption would have captured — and it's a sharper argument for funding adoption than any slide McKinsey could hand you.
Grounded in Lamarre, Smaje & Zemmel, Rewired (McKinsey), and McKinsey's research on why gen-AI pilots stall (A generative AI reset, The state of AI).