Facility-scoped access
Orosyl keeps workspace access and operational views bound to the authorized facility and role.
Practical trust for a controlled pilot
Trust is a workflow, not a badge.
Orosyl is designed to make operating responsibility, evidence, and boundaries visible. The controlled-pilot approach is deliberately narrow so a facility can evaluate the workflow without surrendering authority over data or clinical decisions.
Core operating controls
What Orosyl is built to make visible.
Attributable workflow
Follow-up, escalation, rationale, and closure activity are designed to be visible to accountable facility teams.
Read-only operating posture
The controlled-pilot model focuses on approved source references and does not replace the clinical record or introduce unapproved write-back.
Human decision ownership
Orosyl supports work coordination and evidence review; facility professionals retain clinical and operational judgment.
AI governance foundation
Off by default. Governed before it is useful.
The planned Govryn integration is a policy-evaluation layer for a future, narrowly bounded documentation-evidence assistant. It is not an autonomous clinical tool and it is not active for production facility data.
Current state: the server-side adapter and fail-closed controls are present; live use requires a separate, approved sandbox configuration and documented facility decision.
Guardrails before activation
- The Govryn pilot adapter is disabled by default and requires explicit server-side configuration.
- The initial AI pilot is constrained to de-identified documentation-evidence checklist content.
- Potential identifying or clinical-source content is blocked locally before an outbound policy call.
- A block, timeout, invalid decision, or unavailable policy service fails closed and returns control to the human workflow.
Pilot review packet
Bring the right owners into the first conversation.
The controlled-pilot discussion covers workflow fit, role map, source-reference boundary, decision rights, review cadence, and the documented exit decision. It does not require an EHR write-back or production AI activation.