AI & privacy
AI with clear boundaries.
Care in your hands.
You decide what Holi reads, what enters the patient record and when a care state is approved. Consent and source evidence stay part of the work.
How the workflow works
You decide what AI can use.
Permission to read, approval of a record and approval of a care state are separate decisions. Each gives you a point to pause, check and choose.
Permission to read
Uploading a file and asking Holi to read it are separate choices. You decide what information enters that AI workflow.
Approval of the record
Review the structured record against the source. Approve it before it becomes part of the patient record.
Approval of the care state
Choose which records inform the next version. Review the clinical question, working hypotheses, safety flags and citations before you approve it to go live.
Avery Example · Approved patient records
Records for the next update
3 records selected
A care state is a versioned overview of the patient, built from records you select. The next version needs your review and approval.
A history you can review
Follow the decision back to its source.
Session findings link back to the evidence. Care states keep their version history. File logs include rejected items, so you can review how the clinical picture developed.
Previous care state
Source linked
Approved patient overview
Selected records · practitioner-approved
Data handling
Understand how your data is processed.
On-device transcription and cloud AI serve different parts of the workflow. The distinction matters when you choose how to work.
- On your computer
Choose on-device transcription for live consultations. Your computer processes the audio for transcription in this mode. Online transcription is also available as a separate choice.
- Cloud processing
Session reports, reading lab results, care states and Holi use cloud processing. Your choices about sources and record approval remain part of that workflow.
Before introducing patient information, talk with us about current hosting, processing providers, retention, and access or deletion requests. We'll explain the arrangements for the product and how they apply to your practice.
Ask about data handlingPrivacy noticeBuilt for the profession
Built around professional responsibility.
Consent, practitioner review and traceable sources are central to our approach. AI can be wrong. Check the source, assess the context and use your clinical judgement.
Designed in line with the NHAA's policy on AI in practice.
This describes our design approach. It does not imply NHAA endorsement.