L03 Data
Data before model behavior
A model can only reason over what has been structured, linked, governed, retrieved, and trusted enough to use.
A lot of AI discussions jump straight to the model, but practical systems usually fail earlier. The data is messy, split across tools, missing identifiers, or not trusted enough to automate against.
Before an AI system can answer a question, someone has to decide what the records mean, how they connect, where they came from, and what confidence is acceptable.
That is why formatting, schema design, traceability, and validation are not boring prep work. They are the foundation that makes the model useful.