Safety culture is the best predictor of clinical AI adoption
Units where people already speak up absorb new tools faster and more safely.
5 min read
The link nobody plans for
Units with a strong speak-up culture surface AI problems early, fix the workflow, and move on. Units without it work around the tool in silence.
The second pattern looks like adoption in the data and is the more dangerous of the two.
What to measure
Ethical reasoning and judgment scores, taken together, give a reliable read on whether a unit will raise an issue or absorb it quietly.
Both are measurable with short behavioral scenarios rather than surveys about how people feel.
Acting on a low score
The intervention is rarely more training. It is usually a clearer escalation path and a visible example of a concern being raised and resolved.
Score the unit again next cycle and you will see whether the message landed.