Trust Loop
Trust Loop is the operating pattern that holds together all of nexmin's clinical intelligence: the AI proposes, the therapist decides. No draft generated by the AI — not a note, not a synthesis, not a score, not a proposed intervention — enters the chart without the clinician's explicit approval. It is not a configurable option: it is a design constraint.
The name comes from the logic it articulates: the AI produces a draft, the clinician reviews it, edits it if needed and approves it. Only then does the content publish to the clinical record. If the clinician does not approve, the draft is discarded. The AI learns from the clinician's style through the corrections applied, without that learning retroactively contaminating what has already been published. The reason is epistemological before it is technical. The boundary between observing and interpreting is the professional's responsibility, not the model's. A system that publishes directly what the AI infers erases that boundary and turns the tool into a quasi-autonomous clinician — exactly what the two 2026 AEPD (Spain's data protection authority) notes on AI transcription ask to avoid through "effective human oversight". The Trust Loop materialises that principle: there is a human in the loop, always, with no exceptions. In daily practice, the Trust Loop adds between 3 and 8 minutes of review per analysed session. That time is already accounted for in the honest savings calculations we publish. If at some point someone offers you clinical AI "with no review needed", be clear about what they are selling: a tool that takes clinical decisions on your behalf.
Inside nexmin
Every nexmin AI engine (Scriba, Pensa, Cartógrafo, Phenotype) hands over drafts that you review before they enter the chart. The review interface is built to be quick — most of the time reading and approving is enough — but you are always the one who hits the button.
Related terms
Last updated: 2026-06-11