AI Transparency & Model Behaviour Policy
How AI decisions are made and communicated to clinicians.
Version: 1.0
Effective Date: 9 August 2026
Issued by: Respocare (Pty) Ltd
1. Purpose
This policy explains how artificial intelligence operates within the Respocare Connect AI platform.
The objective is to ensure transparency regarding:
- how AI processes clinical information
- what AI outputs represent
- the limitations of AI systems
- how clinicians remain in control
Respocare believes that trust in healthcare AI must be built through transparency, accountability, and responsible system design.
2. Role of Artificial Intelligence in the Platform
Artificial intelligence within Respocare Connect AI is designed to assist clinicians with clinical information processing.
The system may support tasks including:
- summarising patient records
- organising clinical documentation
- transcribing dictated clinical encounters
- retrieving relevant clinical information
- generating structured documentation drafts
The system does not replace clinical judgement.
3. Human Oversight
Respocare Connect AI is designed around a human-in-the-loop model.
This means:
- clinicians remain the final decision-makers
- AI outputs require clinician review
- clinicians approve documentation before it is stored
The system supports clinical workflows but does not act independently.
4. AI Behaviour Model
The AI system operates through structured computational processes that analyse available clinical data.
AI responses may be generated by:
- analysing patient records
- retrieving relevant information from stored documents
- interpreting clinician prompts
Outputs are generated using probabilistic reasoning models trained on large datasets and configured to operate within the constraints of the platform.
5. Sources of Information Used by AI
AI responses within Respocare Connect AI are generated using information from:
- clinician-provided prompts
- patient records available within the platform
- uploaded clinical documents
The system does not independently access external medical records or patient databases.
All responses are limited to the data available within the platform environment.
6. Limitations of AI Systems
Users must understand that artificial intelligence systems have inherent limitations.
AI may:
- produce incomplete summaries
- misinterpret complex clinical context
- omit relevant information where records are incomplete
AI systems should therefore be treated as assistive analytical tools rather than authoritative sources of medical judgement.
7. AI Output Verification
All AI-generated outputs must be reviewed by the clinician before being used in clinical care.
Examples include:
- reviewing generated clinical notes
- confirming patient summaries
- verifying extracted medical information
The clinician remains responsible for confirming the accuracy and relevance of any generated output.
8. AI Training and Data Usage
Respocare does not use patient data from the platform to train public AI models.
Clinical data processed within Respocare Connect AI is used solely for the purpose of operating the platform and supporting clinician workflows.
Respocare maintains strict controls to ensure that patient information is not reused or repurposed without appropriate authorization.
9. Bias and Ethical Considerations
Respocare recognises that AI systems can be affected by bias present in training data or clinical documentation.
To mitigate these risks, the platform is designed with safeguards including:
- clinician verification of outputs
- transparency of information sources
- restricted decision authority for AI systems
Clinical responsibility always remains with licensed healthcare professionals.
10. Continuous Improvement
Respocare monitors system behaviour and user feedback to improve the performance and safety of the platform.
Improvements may include:
- refinement of AI prompts and workflows
- improvements to clinical summarisation accuracy
- enhancements to documentation tools
These updates are implemented in accordance with responsible AI development practices.
11. Ethical AI Principles
Respocare Connect AI is developed according to the following principles:
Transparency — Users must understand how AI functions.
Human oversight — Clinicians retain authority.
Safety by design — The system operates within clear limits.
Data protection — Patient information is protected.
Accountability — Clinical responsibility remains with healthcare professionals.
12. Summary
Respocare Connect AI is designed to provide assistive intelligence for clinicians, not automated medical decision-making.
The system operates within controlled boundaries to support clinical workflows while maintaining professional responsibility with the treating clinician.