The AI Transformation Pyramid
Why most AI initiatives fail, and the five-layer order that actually works.
Most companies are doing AI transformation backwards. They start with a hackathon.
Someone at the top decides the organisation needs to "do something about AI," so they run an event, or stand up a centre of excellence, and wait for value to appear. It rarely does. Not because the people are wrong or the tools aren't good enough, but because the foundation underneath isn't there yet.
AI transformation isn't one decision. It's a stack of decisions, and each one depends on the layer below it being solid.
That stack is a pyramid. Five layers, bottom to top. You can't skip one and expect the ones above it to hold.
// value falls out the top; everything below it is the foundation
1 · Budget
Someone has to decide how much the company is willing to spend, per employee, per engineer, and per product, before anything else happens. Without a number, every AI decision downstream turns into a negotiation. With one, teams know what they're allowed to try.
2 · Tools & Security
Budget without tools is just a line item. But tools without security is a liability. This layer is about giving people real, sanctioned access to good tools, and doing it in a way that keeps your data and your customers safe. Get both halves or you get neither.
3 · Culture
Culture is where "we have access to Claude" turns into "everyone uses Claude as a default, every day." Access is not adoption. This is the layer of expectations and incentives: the habits that make the tools part of how the work actually gets done, not a tab nobody opens.
4 · Experimentation
Tools are what you give people. Experimentation is what people actually do with them. This is the layer where individuals start finding the automations that matter to their own jobs: the small, specific wins that a central team, sitting far from the work, would never think to look for.
5 · Value
Value isn't a layer you build. It's what falls out the top when the other four are working. You don't engineer it directly. You build the foundation, and value is the output.
Companies that try to engineer value directly, without the foundation underneath, almost always produce theatre instead of results.
The central-team trap
There's a common variant of this failure worth calling out: the dedicated central AI team that everyone routes automation requests through. It feels organised, but it's a separate failure mode. That team lacks proximity to the actual problems, and it moves too slowly. The people closest to the work are the ones who should be experimenting. The pyramid is designed to put the tools in their hands, not in a queue.
Start at the bottom. Set the budget. Get the tools and the security right. Build the culture. Let people experiment. The value takes care of itself.
The five layers
§ recapDecide the spend first
Per employee, per engineer, per product, before anything else happens.
Access, made safe
Budget without tools is a line item; tools without security is a liability.
Default, every day
Turn "we have access" into a habit the whole team actually runs on.
Find the real automations
The people closest to the work spot the wins a central team never would.
The output, not the goal
Build the foundation and value falls out the top on its own.
Want help building the foundation?
// I help teams get the order right, without the hype