Sign in
AI Voice Contact Center

A voice agent that answers member calls — governed.

Production architecture, not a chatbot: the model decides intent, your system of record decides truth, and every turn lands in the audit log. Portable across any model and any cloud — built for ACA, Medicare, and ICHRA call volumes.

REGION  in your tenantFACT FIDELITY  verbatimAUDIT  per-turnPOSTURE  fail-closed
Architecture

One call, end to end — on infrastructure you can audit.

CALLER800 number · webYOUR SYSTEMSCisco · Genesys · ConnectINTENTmodel plane · any LLMGOVERNED FULFILLMENTtyped tools · verifiedSYSTEM OF RECORDyour database · RAGTHE MODEL DECIDES INTENT · THE DATABASE DECIDES TRUTH · EVERY TURN LOGGED

System of record, not a script

Your database — any Postgres-class store — holds members, addresses with history, dependents, claims, payments, and a per-turn audit log. Every account fact the agent speaks is read verbatim from the record — the model never improvises data.

Through your stack, not around it

Calls arrive over your existing systems — Cisco, Genesys, Amazon Connect — into the intent plane, where any model decides what the caller wants. Verification is enforced in the playbook, fulfillment runs in a private per-tenant service, scoped, fail-closed. Any cloud, with or without Dialogflow.

Compliance is the spine

Identity verification before any account fact. RAG-grounded plan documents so copay answers cite a source. Per-action audit trail shipped with the call — the same licensed/unlicensed separation discipline as the rest of HealthNext.

What a pilot is worthhonest math

Run your own numbers.

Practitioner data, not a vendor’s 90% slide. Every figure ranges from the conservative case to the optimistic one, so the floor is a number you can defend in a board review.

50,000
$10.00
60%

30,000 eligible calls/mo · AI runs $0.20–1.00 each · savings net of AI cost

Year one · 20–35%conservative start
Calls deflected / month
6,000–10,500
Annual net savings
$648,000–$1,234,800
Year two · 30–45%eval-driven maturity
Calls deflected / month
9,000–13,500
Annual net savings
$972,000–$1,587,600

We plan to the conservative number. A 15% deflection plan that reaches 30% by year two beats a 40% opening assumption — and it’s the one that survives the board review. Containment compounds as the eval suite tightens; it doesn’t arrive on day one.

Get the full model + a pilot plan
Scripted call scenarioshealthcare lines

Watch it handle claims, Medicare, enrollment, and provider lines.

Live demo

Talk to Eva. Verify a member. Watch the database answer.

The demo runs against the real backend: verify as a seeded member, ask about coverage, change an address — it persists. Synthetic data, real plumbing. Verify as Sam Steele (June 30th, 1971), Maria Alvarez (March 14th, 1985), or David Chen (November 2nd, 1978).

Tip: ask Nova (bottom right) to run the demo for you — she’ll hand you over to Eva with a member identity ready.

Multi-client operations

Run it for your clients — each one isolated, each one auditable.

For service providers and TPAs operating member services on behalf of payer-scale clients. The per-tenant architecture is already in the stack: agents, fulfillment scopes, document stores, and audit logs are provisioned per client from setup-as-code.

Tenant isolation

Per-client agents, scoped fulfillment, separate RAG document stores. One client's change never touches another's.

Supervisor tools

Monitor live calls, whisper to the agent, take over the conversation — transcripts and dispositions land on the determination record with the audit trail.

Compliance evidence

HIPAA-aligned posture, BAA-ready engagements, per-turn audit logs, verification-before-facts enforced in the playbook — evidence packs per client, per period.

Build your own agent. Talk to it in 60 seconds.

Name it, pick a voice, give it a job and your FAQs — then have a live conversation. Free trial, no signup. Production agents run on your systems with verification and audit trails.

Open the agent builder