Formatting, summaries and suggestions
AI can structure a note, produce a summary and suggest wording. The suggestions are drafts that the user edits or discards.
Trust Center
How FlowDule uses AI
AI is optional documentation support in FlowDule, not a decision maker. Here you can read what AI may help with, how people stay in control, and which providers are involved.
AI helps with the writing. People decide.
FlowDule offers AI as optional documentation support that saves time on written work. AI produces drafts and makes no professional decisions about clients, treatment or operations.
This page gathers what AI may help with, who processes data along the way, and how you decide yourselves whether a feature is activated.
Use
What AI may help with
After separate activation: writing support, report drafts, translation, recognition of images and speech, and encryption of AI data
Report drafts
AI can draft a report from what is already written in the journal. The draft has to be approved or rejected before it can be used.
Translation between languages
AI can translate content between the platform's languages, so text can be read by the person who needs it.
Text in images and speech to text
AI can read text in images and documents (OCR) and write speech out as text, so content does not have to be typed in manually.
AI content is encrypted like other journal content
Input to and output from the AI features is encrypted at rest in the same way as the rest of the journal content.
Control
A person approves before anything becomes part of the journal
Generative output, OCR, transcription and translation are only entered into the journal once a trained, responsible user has actively compared them with the source and approved them. There is no automatic completion or signing of records without active human approval. Access is tied to completed training, and every use case must pass a quality test with error thresholds and stop criteria before it is released for client, journal and health data. AI is optional, and records can always be kept without AI.
AI information
Providers
AI processing takes place with AWS, and the regions are documented per feature
The planned provider chain uses AWS services: Amazon Bedrock with an approved model, currently planned as Anthropic Claude via AWS, plus Transcribe, Textract and Translate. Bedrock uses an EU inference profile, and the specific destinations are still open. The actual model versions, AWS regions and any inference destinations may change; the binding combination follows from the accepted version of the sub-processor list and the customer's instruction annex. Model, destinations, retention mode, allowed modes and provider sharing must be documented before activation, and the customer's client and journal data must not be used for model training.
Sub-processorsQuestions?
Questions about our use of AI?
Write to privacy@flowdule.com if you need more detail about the AI features, the providers or how the features are activated and switched off.