Technical evaluators and data protection officers
Data governance
A model is chosen for a task, and each task states what data it is allowed to read.
Task to scope
| Task | Data it may read |
|---|---|
| Note structuring | The current encounter and the linked history |
| Coding proposal | The signed note and the payer rule set |
| Message drafting | Appointment and plan data, no free clinical text |
| Queue prioritisation | Operational metadata only |
How much is AI, exactly
A precise share, not a slogan: this is the count of capabilities where a model proposes and a person decides.
16%
343 of 2,156 capabilities are AI-augmented
- Clinical Operations26%
- Workforce Operations13%
- Governance & Compliance9%
- Analytics & Performance6%
- Growth & Access5%
- Revenue & Finance4%
- Platform & Interoperability2%
- Public Health & Community0%
The six questions
- What is it?
- A routing table from tasks to models, with the data scope each task may read.
- Who is it for?
- Technical evaluators, data protection officers and security reviewers.
- Which problem does it solve?
- A single general-purpose model for everything is either too weak for structured clinical work or too permissive with data.
- How does it work?
- Structuring, coding, summarising and routing are separate tasks with separate models and separate scopes. Customer records are not used to train models. Processing region is a configuration, not an assumption.
- How is it different?
- Model choice is an operational setting that can change without changing the workflow around it.
- What is the proof?
- Each automated task can report which model handled it and what scope it read.
This page is part of a platform of 2,156 clinical and operational capabilities across 8 business domains.
Explore the catalogueContinue
A model is chosen for a task, and each task states what data it is allowed to read.
Legal review: 2026-08-24