Walk the floors of one large organization and you can visit two different decades. On the executive floor sits an AI-native CTO: agents drafting, routing, summarizing, half the job running through machines. A few floors down sits an analyst working exactly the way analysts worked in 1999: spreadsheets, manual pulls, copy, paste, repeat. Same company. Same tools on offer. Decades apart.
BCG has a name for this: the Silicon Ceiling. In their figures, AI adoption among leadership runs around 85% while frontline adoption stalls at roughly 51%. The ceiling is that gap itself: the distance between what the top of the organization has embraced and what the front line actually uses.
The old glass ceiling blocked people from rising. The silicon ceiling blocks capability from descending. Adoption pours in at the penthouse and stops partway down the building, well before it reaches the floors where most of the work, and most of the headcount, actually live.
Here is my inference about why the gap matters more than any single adoption number. An organization stuck at a uniform 60% has a diffusion problem; you fix it with access and training. An organization at 85-over-51 has a translation problem: the layer that decides has changed how it works, while the layer that executes has not. Every AI strategy written above the ceiling is describing an organization that doesn’t exist below it.
Two ceilings, one building
Readers of the Stage 2 Ceiling (opens in a new tab) will recognize the shape. That ceiling is individual: people get faster, but the organization doesn’t change. The Silicon Ceiling is the same failure at the level of organizational strata: leadership gets fluent, and fluency never travels down. One caps the value of adoption; the other caps its reach. And they compound. A leadership team comfortably stuck at Stage 2 of the maturity journey (opens in a new tab), sitting above a front line that never cleared Stage 1, will show you a healthy average adoption number and a business that hasn’t moved.
Finding your own ceiling
You will not see this gap in a company-wide adoption metric; the average hides the strata. My recommendation: measure adoption by altitude. Take your leadership number and your frontline number, put them side by side, and treat the spread, not either figure, as the health of the program.
What to take into the room: ask for AI adoption reported as two numbers, top and front line. If yours look anything like BCG’s 85 and 51, you don’t yet have a transformation. You have an executive perk. The work is not to celebrate the top floor. It is to move capability through the ceiling, floor by floor, until the analyst and the CTO are working in the same decade.