A visible clinical boundary
Cairava adapts how the hospital-approved plan is executed, not what medical care should be delivered. AI supports language and workflow, approved rules control safety routing, and licensed people make clinical decisions.
0 diagnoses · 0 dose changes · 0 AI-only clinical decisions
The boundary in practice
Workflow automation and approved rules can move information toward review. They cannot interpret it or close the clinical concern on their own.
✓Explain approved plan content in plain language
✓Organize transition obligations and due windows
✓Adapt communication and task sequencing
✓Capture patient-reported barriers
✓Apply hospital-approved routing and response rules
Every event remains traceable—Diagnose, interpret symptoms, or determine clinical severity
—Recommend treatment or change a medication
—Set or change escalation thresholds and pathway rules
—Resolve a clinical conflict
—Close a clinical concern without documented human action
Owned by licensed care teamsProvenance stays visible
Every important item should show where it came from, whether a person has reviewed it, and what the system is allowed to do next.
Authenticated server boundary
The patient experience does not write directly to clinical tables or resolve conflicts locally.
A person intentionally submits the reported update.
Authorization and scope are checked before mutation.
A stable key prevents an accidental duplicate action.
Clinical meaning and escalation remain staff-owned.
Source, owner, and resolution remain traceable.
Privacy-conscious by construction
These are design and engineering principles, not a certification claim. Production controls remain subject to hospital security, privacy, legal, and deployment review.
Each surface and service receives only the information required for its narrow role.
Notification titles and bodies remain generic; sensitive context belongs behind the authenticated app boundary.
Permanent secrets remain on the server. Client experiences receive only short-lived, narrowly scoped access.
Names, symptoms, medications, measurements, appointments, free text, tokens, and discharge documents are excluded from logs.
Questions, clearly answered
No. AI supports language and workflow, while hospital-approved deterministic rules control routing and response deadlines. Diagnosis, symptom interpretation, treatment, medication changes, and clinical resolution remain with licensed care teams.
Patient-reported information remains visibly distinct from verified clinical information and pending review state. It is not silently promoted into the verified plan.
No. Clinical mutations must cross an authenticated server boundary, carry an idempotency key, and follow the hospital-approved workflow.
No. It describes Cairava’s intended product and engineering boundaries. Contractual, regulatory, security, and deployment claims require a separate review for each production setting.
Make the boundary reviewable
A walkthrough covers what the system prepares, where it stops, and what the hospital team must approve before any deployment.