Keep your attention in the conversation.
Planned consented speech-to-draft support, with clinical review and approval built into the record workflow.
Explore the story
Medrella roadmap. This experience is being planned; scope and readiness will be demonstrated before implementation.
What this workflow connects.
The scope below guides discovery and demonstration. This capability is on the Medrella roadmap.
- 01
Consented conversation or dictation capture
- 02
Structured draft notes with encounter context
- 03
Clinician correction and approval
- 04
A manual path when assistance is unsuitable
Documentation should follow the encounter
The consultation is a human conversation, while the clinical record needs a clear and attributable account. Moving between those modes can create clerical work during and after the visit.
Medrella's planned documentation assistant aims to prepare a structured draft from consented speech or dictation so that the clinician can review it in context. The objective is to reduce unnecessary transcription effort without weakening professional responsibility. A draft is not a final note, and fluent language is not evidence that the content accurately represents what was said or clinically intended.
Capture only the context the workflow needs
The assistant should operate within an explicit encounter and a defined consent process. It must distinguish speakers where that matters and avoid turning background conversation into a clinical assertion.
The source, timing and draft status should remain visible to the reviewer. Different specialties may need different structures, and the clinician should be able to use a narrative where a template is inappropriate. Audio retention and access require deliberate decisions. The product should not retain every recording by default simply because storage is available or a later feature might find it useful.
Make uncertainty easy to review
A useful draft highlights what requires attention rather than presenting every extracted detail with equal confidence. Names, medicines, negations and timing can change the meaning of a note.
The reviewer needs a practical way to compare the draft with its source and correct omissions or unsupported statements. The assistant should not infer a diagnosis or treatment decision that the clinician did not make. Editing, rejection and approval should be ordinary parts of the workflow. Once approved, the record needs the same attribution and amendment discipline as documentation created without AI.
Connect the note without granting hidden authority
Documentation may mention an investigation, a medicine or a follow-up intention, but that mention should not silently become an approved order.
Draft extraction and transactional actions need separate review boundaries. The clinician should know which parts will update the record and which require an additional decision. Any patient-facing summary should be based on approved content. This distinction lets AI assist with clerical preparation while deterministic systems retain responsibility for permissions, identifiers and changes to authoritative records. The integration must be designed around the actual clinical application, not only a standalone transcription demonstration.
Measure correction burden as well as time saved
Evaluation should include different accents, interruptions, specialty language and deliberately ambiguous statements. Ask clinicians to review the output without assuming it is correct.
Measure missing information, unsupported additions and the effort required to approve a usable note. Compare the complete documentation task with the current workflow, including consent and review. A model that produces a quick draft may still create more work overall. Medrella's scribe is planned; production use requires an agreed scope, clinical evaluation and evidence that the review process is practical for the people responsible for the record.
Connected thinking.
What would more time
for care make possible?
Let's explore it together