Care-team support
Surfacing which complex, multi-condition patients need attention, drafting the supporting documentation, and recording why. Output goes to a care team, not to a patient.
Transparency / applied model card
The label on the tin.
Health systems now expect an applied model card before a vendor system touches their patients. Here is ours for the clinical intelligence layer, following the Coalition for Health AI applied model card structure.
Applied model card / clinical intelligence layer CHAI-aligned
Model details
SystemClinical intelligence layer
DeveloperAffinity Networks, Inc.
TypeAgentic system, multiple models
SettingMedicare value-based care
OversightNamed medical director
Validationfigures pending publication
Data & privacy
Update cadence
// amber fields are pending final validation figures; ask for the current card card v1
Intended use
What it does, and what it must not.
An intended-use statement only means something if the out-of-scope list is just as explicit. Both are enforced in the pipeline, not appended as a disclaimer.
In scope
In scope
Documentation with provenance
Every surfaced recommendation carries its evidence trail: the observations, the rules, and the model calls behind it, replayable end to end.
Out of scope
Autonomous clinical decisions
The system does not diagnose, prescribe, or act on a patient. AI drafts, clinicians decide. Final clinical judgment always belongs to a licensed clinician.
Out of scope
Unreviewed patient contact
No output reaches a patient without clinician review. The system has no authority to message, schedule, or bill on its own.
Human oversight
Clinicians hold the pen.
H1
Named medical director
Clinical accountability sits with a named medical director, not a product team.
H2
Clinician sign-off
High-stakes outputs require clinician review before they take effect. The sign-off is recorded in the audit trail.
H3
Full replay
Every decision is traceable and replayable for clinical review, audit, or dispute.
Validation & maintenance
Proven the boring way.
The system passes the evaluation gate described in how we evaluate: golden sets built from real care-team work, adversarial runs, and clinician review before any autonomy is granted. Published validation figures are pending and will replace the amber fields above; interim results are available to customers on request. Post-deployment, the eval suite runs continuously, drift is monitored against the golden set, and every model or policy update passes the same gate before it ships.
Read it before you buy it.
We will send the complete card, current validation status, and the evidence behind any line on this page.
Request the current cardclinical: build@theanigroup.com
standard: CHAI applied model card