What's this about?
dotey's test: look at how a company's processes are designed. Are they built around people doing the work, or around AI agents doing it? The key point is that in an AI-native company the agent is the execution subject, and humans define the problem and accept the result. Adding a bit of AI into your existing way of working is not AI-native. yangyi proposes a complementary metric he calls the "intelligent data ratio": how much of a company's data is actually reachable by an agent. If production, inventory, marketing, sales and support data are scattered — or on paper — the agent has no context, and without context it cannot do the work.
What is the actual claim?
That AI-native is a structural property, not an adoption statistic. You can have every employee using a chatbot daily and still be organised entirely around human execution. The test is what happens when work arrives: does a person do it with assistance, or does an agent do it with a person defining and accepting? And yangyi's addition is the precondition — that transformation is impossible if your data is not digitised and reachable, which for most non-software companies is the actual first step.
Why is it worth noticing?
- Two independent definitions posted days apart, and they compose neatly: one is the process test, the other the data prerequisite.
- Both are falsifiable, which is more than most AI-native discourse manages.
- Read against "good culture beats AI" from this week's HN — one says restructure around agents, the other says the bottleneck was never execution speed.