You are probably building the wrong layer
The only RCT in this space decomposes where the benefit comes from, and the split is not close.
Lesson 02 of 06 · 10 min
167 professional security analysts, a 93-item corpus, three arms: control with no AI, an agent-prioritised queue with visible verdicts and rationale, and the same prioritised queue with no AI references shown at all. That third arm is what makes the decomposition possible.
Of the total productivity gain, how much came from queue prioritisation versus verdict awareness?
The rest of this lesson unlocks when you commit above.