A way of thinking about AI deployment
The ground is baked hard. Processes are brittle, manual, disconnected. It’s not that nothing was ever tried—seeds were planted, water was poured. But the soil couldn’t hold it. Everything dried out or ran off.
This is the state before meaningful AI deployment — hardened by years of conditions that wouldn’t let anything take root.
AI arrives. Early efforts pool in isolated crevices—a chatbot here, a document summarizer there. Most of the water beads on rock and runs off. Some evaporates before it can do anything useful.
Skeptics point at the puddles and declare failure. They’re not wrong about the puddles. They’re wrong about what comes next.
Each effort softens the soil a little. Water seeps into cracks that were sealed shut. Roots—workflows, integrations, changed habits—spread underground where nobody’s looking.
Here’s the test for whether it’s working: the second use case in a domain should cost a fraction of the first. If every new effort starts from zero, the water is still beading.
Enough micro-problems get solved that workflows start linking up. What was isolated becomes networked. A summarizer feeds a routing system feeds an analyst’s dashboard—none of which was planned that way.
Momentum builds not from one big bet, but from a hundred small ones finding each other.
From the hillside you can sometimes glimpse it: a valley downstream, green, fed by something larger than any single channel. That’s the horizon. You orient by it, not toward it.
Nobody builds a river. You create the conditions for one to form—and you never know in advance which crevice was the one that mattered. So you keep filling crevices.
You don’t need to believe the river is coming to justify the rain.
Every crevice filled, every patch of soil softened, every root established—these deliver value on their own. A team that can summarize faster. A process that routes correctly. A decision made with better information.
The rain is worth it even if the river takes longer than you’d like.