iBridge DigitaliBridge Digital
← back

How it works

From your first callwith us to your firstcolleague inproduction.

A Smarna colleague onboards the way a new staff member would: assigned to a workflow, given a phone number, paired with a training authority, and taught by voice. The difference is she begins work on day one and keeps learning every day after.

Engagement

Six to eight weeks from first call
to a colleague in production.

01 / Discovery

Week 1

We map your operational layer.

30-minute call. We map your call volume, payer mix, EHR or RCM stack, the workflows that hurt most, and what good looks like. You leave with a one-pager: which agents we'd deploy, in what order, with what integration depth.

02 / Benchmark

Week 2

We run your real calls through our agents.

You send 100 representative calls (or call recordings, or job samples). We process them through Smarna in a sandbox tenant. You get back: cost per call, accuracy by field, average handle time, and a confidence-scored breakdown of where the agents win and where they hand off.

03 / Pilot

Weeks 3 to 6

One agent. One workflow. One clinic location.

We deploy a single agent against a bounded scope: e.g., outbound prior auth status calls for one specialty at one location. Authority chart set. Knowledge sources connected. Voice training begins on day one with your training authority and one or two peers.

04 / Production

Week 7+

Scope expands as the colleague proves herself.

In-scope teachings ship continuously. Out-of-scope changes flow through the architect for approval. Audit log is queryable from day one. We add agents and locations as your team is comfortable. No big-bang cutovers.

What makes a colleague different from a bot

Four design principles.
All four show up in every agent we ship.

Voice training, not config UIs

A peer dials her number and teaches her by voice. She listens, reasons, asks a clarifier if needed, pushes back politely on contradictions, reflects back, and announces routing transparently. No prompt-editing dashboards for clinics.

Authority and scope, by tier

Owners, training authorities, peers, architects. In-scope changes apply. Out-of-scope changes create an escalation flag with the full reasoning chain. The colleague refuses what her clinic hasn't authorized.

Knowledge in layers

Foundation knowledge from the LLM. Authoritative APIs (CMS, AAPC, payer policy) consulted live. Clinic-specific semantic memory from voice teaching. Risk discipline scaffolded into every decision.

PHI never reaches the core

PHI lives in process memory for the duration of one call, then is destroyed with a hashed proof. The colleague's reasoning, persisted memory, audit log, and customer-facing surfaces never see PHI by architectural enforcement.

Integration

We meet you
where your data lives.

EHR, PM, RCM, telephony, payer APIs. We integrate at the level appropriate to the engagement — from API-first deep integrations to portal-orchestration to flat-file batch.

EHR / PM

  • Epic
  • athenahealth
  • eClinicalWorks
  • NextGen
  • Greenway
  • DrChrono

RCM platforms

  • Waystar
  • Availity
  • Change / Optum
  • AdvancedMD
  • Kareo

Telephony

  • Twilio
  • RingCentral
  • 8x8
  • Five9 (SIP)

Knowledge

  • CMS NCD/LCD
  • AAPC
  • Payer policy APIs
  • FDA NDC

What you see

Four customer-facing surfaces.
All of them read-mostly.

Smarna doesn't hide behind dashboards full of knobs. Your team reads what the colleague is doing, what she's learning, what's waiting on review, and what's been flagged. Approval is the only write action.

Call Reader

Every call the colleague makes or takes, turn-by-turn, with detected issues highlighted and structured outputs attached.

Training Inbox

Pending peer teachings awaiting verification by the training authority. One-click verify, correct, or reject. Or review by voice.

Architect Inbox

Out-of-scope escalations with the full reasoning chain. Each flag shows what change, why out-of-scope, what knowledge was consulted, recommended routing.

Risk Register

Open risks the colleague has identified, with severity, likelihood, recurrence count, status, and recommended response.

Next step

Six to eight weeks. One agent.
Bring us the workflow that hurts most.