Everything this observatory measures about AI evaluators generalizes to "commercial evaluation" only through a bridge: resemblance, influence, or replacement. These are three different research programs with three different evidence requirements. The discipline's first job is testing its own premise — which is why this question is written down before any finding exists.
- Affected buyers
- Buying committees whose members consult AI systems during evaluation
- Affected categories
- Categories where AI-assisted screening is becoming standard practice
- Potential product impact
- Determines whether optimizing for AI evaluators matters at all — and how much
- Current confidence
- Unresolved. The bridge claim is the program's foundational assumption, held as a hypothesis.
- Evidence tier
- Narrated (untiered type)