• 

• 

Klarity Voice AI Concierge

24/7 after-hours voice intake for prospective patients.

Klarity Voice AI Concierge replaces traditional support-desk intake with an AI phone agent that answers after-hours calls, gathers structured intake context, screens for safety concerns, matches an appropriate provider, reserves a mock appointment, and prepares a provider-ready summary.

Built for the ** Retell + Insforge + Zeabur applied to Klarity Health.

---

Why This Matters

Mental-health clinics frequently miss prospective patients after hours because front desks are closed, intake is manual, and providers receive fragmented information.

This project demonstrates how AI can improve intake operations while maintaining healthcare safety guardrails.

For Patients

• 24/7 after-hours support
• Faster scheduling experience
• Reduced intake friction
• Faster access to the best next step

For Clinics

• Capture missed after-hours calls
• Reduce front-desk/admin burden
• Standardize intake collection
• Improve operational efficiency

For Providers

Instead of reviewing fragmented call notes, providers receive a structured intake summary including:

• presenting concerns
• safety screening results
• appointment readiness
• provider matching
• visit preparation checklist

Demo outcome: A caller goes from phone call → intake → provider match → appointment reservation → provider-ready summary in minutes.

---

Demo at a Glance

Scenario

After-hours mental-health intake

Caller: Maya Johnson (California)

Concern: Trouble focusing and anxiety affecting work

Urgency: Routine

Safety status: No immediate danger disclosed

End-to-End Flow

1. Retell AI answers the phone call
2. Voice agent gathers intake details
3. Structured intake is stored in Insforge
4. Safety/risk screening is completed
5. Appropriate provider is matched
6. Mock appointment is reserved
7. Provider-ready summary is generated
8. Dashboard updates in real time

---

Clinical Safety Guardrails

This system supports intake and routing only.

The voice agent:

✅ Collects structured intake context
✅ Screens for immediate safety concerns
✅ Suggests the best next step
✅ Helps reserve appointments

The voice agent does not:

❌ Diagnose conditions
❌ Recommend treatment plans
❌ Recommend medication or therapy
❌ Replace provider judgment

Approved phrasing

“Based on what you shared, a good next step may be an initial evaluation.”

Avoided phrasing

🚫 “You probably have ADHD”
🚫 “Medication is the best option”
🚫 “Therapy is what you need”

The system is designed to support provider workflows, not clinical decision-making.

---

Architecture

The system combines a phone-based AI intake experience with structured backend workflows and a provider-facing dashboard.

flowchart TD
  Patient["Prospective Patient"] -->|"Phone call"| Retell["Retell Agent<br/>Voice, conversation flow, function calls"]

  Retell --> Create["create_intake_record"]
  Retell --> Match["match_provider"]
  Retell --> Reserve["reserve_appointment"]

  Create --> Backend["Insforge Backend<br/>Patients, Calls, Providers, Appointments, Escalations"]
  Match --> Backend
  Reserve --> Backend

  Backend --> Dashboard["Next.js Dashboard<br/>Provider/admin view"]

  Dashboard --> Zeabur["Deployed on Zeabur"]

Repository Layout

apps/
  api/                 Retell webhooks and custom function endpoints
  web/                 Next.js provider/admin dashboard
packages/
  shared/              Shared TypeScript contracts and demo fixtures
retell/
  agent/               Agent prompt and flow notes
  tools/               Custom function schemas for Retell
docs/                  Architecture, Insforge, Zeabur, and demo guides
insforge/              SQL schema and seed data for the demo backend

Built With

• insforge
• retell
• zeabur