AI AgentHealthcare

MediVoice
Care that answers.

A bilingual, 24/7 patient intake and scheduling agent that answers every call and web chat, books directly into the EHR, sends reminders, and escalates urgent symptoms to clinical staff.

US HealthcareHIPAA-aware12 clinic locationsVoice + Web Chat
Client type12-location outpatient clinic group
Timeline16 weeks from discovery to launch
DeploymentAWS · Voice · Chat · EHR
Top result−72% missed calls after hours
The Challenge

Patient demand did not end when the front desk closed.

Bright Ridge Family Clinics serves roughly 90,000 patient visits each year across 12 locations in Texas and Oklahoma. Its phone system and front-desk workflows, however, were built around office hours. When patients called at night or during peak periods, they often reached voicemail or waited too long to speak with someone.

The same team responsible for welcoming patients in person was spending about 4.2 hours per day, per location, answering repetitive scheduling calls. Manual symptom routing varied by staff member, while inconsistent reminders contributed to a 22% no-show rate.

The clinic needed dependable coverage without putting clinical judgment inside a general-purpose language model or forcing staff to replace the systems they already knew.

38%

After-hours calls went unanswered

Patients who were ready to schedule reached voicemail instead, creating avoidable frustration and lost bookings.

4.2h

Consumed by repetitive calls each day

Every location lost almost half a staff day to booking, rescheduling, confirmation, and policy questions.

22%

Baseline appointment no-show rate

Reminder gaps and language barriers left schedule capacity unused and delayed care for other patients.

The Solution

One patient-access agent, connected to the clinic’s existing workflow.

DEVginix built MediVoice as a voice-and-chat front door that works alongside the clinic’s staff, EHR calendar, policies, and on-call procedures.

For routine requests, the agent verifies a patient, checks live availability, books or reschedules the visit, and confirms it by SMS. For clinical language, it follows a clinician-approved triage tree. Red-flag symptoms trigger an immediate handoff to the on-call team—typically in under 30 seconds.

Protected health information is passed only to approved tools and is not stored in LLM context. Weekly transcript sampling, exception reviews, and policy updates keep the workflow accountable to human owners.

  • Autonomous for scheduling, rescheduling, reminders, and routine policy answers
  • Human-in-the-loop for red-flag symptoms, exceptions, and clinical decisions
  • English and Spanish across phone calls, web chat, and SMS follow-up
Practice Manager Dashboard

Every call, action, and exception in one view.

Operations leaders can see demand across 12 locations without interrupting the front desk or opening individual transcripts.

How It Works

From first contact to completed action.

A controlled five-stage workflow separates conversational understanding, approved knowledge, system actions, and human review.

01

Channel Input

Twilio voice or web chat captures the request.

02

Intent & Entities

The LLM interprets language inside firm guardrails.

03

Knowledge & Tools

EHR, clinic policies, and the approved triage tree.

04

Action & Handoff

Book, reschedule, escalate, or send confirmation.

05

Learning Loop

Weekly quality review informs prompt and tool updates.

Step 4 · Clinician Handoff
Live Interaction

One patient-access agent, two channels.

Sarah books through web chat while James calls after hours. Both receive the same guardrailed workflow, live scheduling access, and documented outcome.

AI
MediVoiceOnline · Handling calls
02:41
Tuesday · 8:43 PM CT
AI
Thank you for calling Bright Ridge Family Clinics. I’m MediVoice, the automated patient-access assistant. How can I help tonight?
SM
Hi, I’m Sarah Mitchell. I’ve had a persistent cough and fever since Saturday, and I’d like to see someone.
AI
I can help with that. Before I access your schedule, please confirm the four digits for the month and day of your date of birth.
SM
Zero seven one four.
AI
Thank you, Sarah. I found your profile and North Austin is your nearest location. Are you having severe trouble breathing, chest pain, confusion, or blue lips right now?
SM
No, none of those. What is the earliest appointment available?
Checked calendar · 3 slots available this week
AI
I have Wednesday at 10:15 AM with Dr. Emily Dawson, Wednesday at 3:40 PM with James Carter, PA-C, or Thursday at 9:00 AM with Dr. Dawson. Which would you prefer?
SM
Wednesday at 10:15 with Dr. Dawson, please.
Booked appointment #A-48210 · SMS confirmation sent
AI
You’re confirmed for Wednesday at 10:15 AM with Dr. Emily Dawson at Bright Ridge North Austin. I sent the details by SMS. Please monitor your symptoms, stay hydrated, and follow the clinic’s fever guidance. If your breathing worsens, you develop chest pain, or you feel severely unwell, call 911 immediately.
MediVoice does not diagnose. Red-flag symptoms follow a clinician-approved escalation pathway and connect to an on-call professional.
J. C. · +1 (512) •••‑4408Incoming after-hours call
Handled by MediVoice
Tuesday · 11:12 PM CT03:42
Patient
Hi, I’m James Carter. I’ve had a migraine most of today and it’s not improving.
MediVoice
I can help check the next appointment and screen for symptoms that need urgent attention. Please confirm the month and day digits of your date of birth.
Patient
Zero four two two. I don’t have weakness or confusion—just the persistent headache and some nausea.
MediVoice
Thank you. The earliest visit is tomorrow at 8:30 AM with Dr. Emily Dawson at the Round Rock clinic. Would you like me to book it?
Patient
Yes, please book that appointment.
MediVoice
You’re confirmed for 8:30 AM. I sent the details by SMS. If the headache becomes sudden and severe, or you develop weakness, confusion, fainting, vision loss, or trouble speaking, seek emergency care now. Clinician-approved emergency guidance delivered
Key Capabilities

Built for real patient-access work.

Each capability maps to a defined clinic workflow, an approved data source, and a clear escalation owner.

24/7 voice and chat intake

Answers simultaneous patient calls and web chats after hours, during lunch, and through peak demand without sending routine requests to voicemail.

Real-time EHR scheduling

Reads provider availability, respects visit types and location rules, then books or reschedules directly in the clinic’s source-of-truth calendar.

Guardrailed symptom triage

Uses a clinician-approved decision tree for safety screening and hands red-flag symptoms to on-call staff within a 30-second target.

English and Spanish

Maintains the patient’s chosen language through voice, chat, appointment details, and SMS reminders without creating a separate workflow.

Smart SMS follow-up

Sends confirmations, preparation instructions, bilingual reminders, and reschedule links using timing rules tied to each appointment type.

Transcripts and operations analytics

Gives practice managers PHI-safe summaries of demand, completion, escalation, transfer, and exception trends across all 12 locations.

Tech Stack

Production architecture, chosen for control.

The stack separates model reasoning from deterministic clinic tools, auditable actions, and PHI-safe operations.

Language & Framework
Python 3.11FastAPI
AI / LLM
OpenAI GPT-4-class modelLangGraphPinecone
Voice / Telephony
Twilio Programmable VoiceTwilio SMSDeepgram STTElevenLabs TTS
Integrations
Athenahealth-style EHR APIGoogle Calendar mirrorSlack on-call alerts
Infrastructure
AWS ECS FargateRDS PostgreSQLAmazon S3DatadogPHI-safe logging
Results

More completed appointments. Less administrative drag.

Measured across the first 90 days after rollout, with operational review by the clinic’s practice-management team.

0%

Missed calls after hours

Unanswered after-hours calls fell from 38% to 11%.

+0%

Booked appointments

Monthly completed bookings increased within the first 90 days.

0h/day

Front-desk time returned

Time reclaimed per location for in-person patient support.

0%

No-show rate

Smart bilingual SMS reminders reduced missed visits.

Results reflect the fictional case-study brief and describe the modeled Bright Ridge Family Clinics deployment.

“Our teams got meaningful time back to focus on the patients in front of them, while callers could finally get help outside office hours. MediVoice gave us dependable coverage without removing staff from the moments where human judgment matters.”
Operations Director · Bright Ridge Family Clinics

What patient workflow should your AI agent handle first?

Bring us the queue, handoff, or access problem. We’ll help you define the workflow, guardrails, integrations, and business case.