Executive Briefing
This document is classified as Confidential.
DME Express Internal Use Only.
Incorrect access code. Please try again.
Confidential — DME Express Internal Use Only
Prepared by Ark Fortune  ·  Under the direction of Anirudh Padiyal

The most common failure mode for enterprise AI projects is scope creep — building a system that tries to do everything and does nothing reliably. This initiative is scoped to exactly two call types that represent the highest daily volume and the lowest clinical complexity.

The AI's intelligence comes entirely from Core System data. Without deep integration into live order records and patient delivery status, this is just a phone bot. Ark Fortune is currently building the DME Reporting Tool and the Odoo Support Dashboard — that active, hands-on experience with DME Express's data model is what makes this proposal uniquely executable.

Pillar 1 — Fully Autonomous
Order Status & Delivery ETA
Read-only Core System access — zero human involvement required

A hospice nurse calls and says: "Where is the hospital bed for Jane Doe?" The AI answers that question — completely, accurately, and instantly — without routing the call to a human agent.

Caller Calls In
Hospice nurse dials the DME line
AI Authenticates
Verifies agency & patient DOB
Queries Core System
Real-time order & delivery data
AI Responds
Speaks ETA in plain language
Call Logged
Auto-logged to Core System. Done.
100%
Autonomous Resolution
24/7
Availability
~3 sec
Response Time
Zero
Hold Time
Pillar 2 — AI-Assisted
New Equipment Order Intake
AI collects — human reviews & approves — no PHI leaves without oversight

A hospice coordinator calls to place a new order: "I have a new patient — they need a hospital bed, a Hoyer lift, and supplemental oxygen." The AI collects every required data point and builds a pre-filled draft order in Core System. The human agent opens a complete ticket and clicks approve.

Order Call
Coordinator places new order
AI Collects Data
Patient info, Dx, equipment, address
Draft Order Created
All fields pre-filled in Core System
Human Reviews
Agent reviews & confirms
Dispatched
Goes to warehouse. Complete.
~50%
Time Reduction per Order
Zero
Data Entry for Agent
100%
Human Approval Required
Full
Audit Trail in Core System

What Always Goes to a Human

The AI is hard-bounded. It does not make clinical decisions. It does not handle exceptions. Any caller or situation outside the two defined workflows triggers immediate, warm transfer to a human agent — with full call context already captured.

Always escalated to a human agent immediately

  • Stat / urgent deliveries
  • Billing disputes & insurance issues
  • Equipment troubleshooting
  • Any caller requesting a person
  • Complex clinical situations
  • Caller distress or confusion
  • Anything outside Order Status or Intake
  • Pickup requests & returns

The Technology Stack

We are not building a voice AI engine from scratch. We are selecting the right enterprise platform, configuring it to speak DME Express's workflows, and connecting it to the Core System data it needs to be useful.

RingCentral (Existing Phone System)
Incoming call arrives on DME's existing number — no number change for callers
Retell AI — Voice Agent Engine (HIPAA-Compliant)
Speech recognition → Natural language understanding → Response generation → Text-to-speech
Core ERP (DME Express Core System)
Pillar 1: Read-only order/delivery query  |  Pillar 2: Create Draft Order (write access)
Human Agent Queue (Escalation Path)
Warm transfer with full call context — agent never starts from zero

HIPAA compliance is architectural, not an add-on. Retell AI is HIPAA-native with a BAA available. All PHI is encrypted in transit (TLS 1.3) and at rest (AES-256). Every AI interaction is immutably logged. The AI accesses only minimum necessary PHI — patient name, DOB, and order data — not full medical records.

Next: Build Plan