A source is raw material and nothing more until it’s processed. A two-hour transcript, a client email, a saved article: each becomes useful only after something extracts, connects, and files the useful parts where future work can find them.
Think about the intake tray on a desk. Mail lands there before anyone decides what it means: a bill, a signed agreement, junk, something that needs a reply. The tray isn’t the finished system. It’s where raw material waits to be sorted.
How it shows up
This is why we separate capture from processing. A script can collect sources all day without AI: meeting transcripts, bookmarks, documents dropped into an intake folder. AI earns its keep when it starts transcript processing, extraction, summarization, routing, and connection. A vault is one common home for processed sources, so the raw transcript stays as evidence while the meeting note, person note, and decision note each carry the useful parts into a clean knowledge base.
Why you care
AI can only reason from what it can see. If the source is missing, the agent guesses. This also connects to RAG, since retrieval only helps when there’s something trustworthy to retrieve. AI work is only as grounded as the raw material underneath it.