1.2M plan-doc embeddings · semantic RAG over your tenant · cross-encoder reranker · plan-grounded decoding · inline citation verifier
Data handling
Amanda learns in aggregate, never from any one person — patterns across the whole book of business sharpen every answer, while your individual record stays sealed. Claims, diagnoses, and medications are never ingested. Encrypted at rest, isolated per employer.
De-identified trainingTenant-isolatedAES-256 at restNo PHI, ever
Plan savings · this year
$0
Surfaced by Amanda steering members to high-value care.
0
questions answered
0%
self-served, no HR ticket
0 hrs
HR time saved / month
0×
point-solution utilization
0%
member satisfaction
0%
lower cost / service
Where the savings came from
Care steerage$92k
Rx savings$41k
Bill errors caught$28k
Program use$23k
Illustrative aggregate across a 480-life group — de-identified, never any one person.