Relay Atelier
All case studies

Client implementation · Hospital pilot

Bringing hospital communication and operations into one layer.

Developed for Özel Yenihayat Hastanesi (Osmaniye, Turkey), Relay Health brings AI-assisted communication, incoming patient requests, conversations handed off to a human operator, and single-point operational visibility into one coordinated system. Instead of replacing the hospital's existing management software, the platform focuses on the communication and operational processes around it.

Product
Relay Health
Sector
Healthcare and hospital operations
Year
2026
Category
Healthcare operations infrastructure

Connection status right now

  • Instagram
  • Facebook Messenger
  • WhatsApp Business
Editorial illustration representing Relay Health's unified communication layer: requests from Instagram and Messenger converging on Pınar before handing off to an operator.
An editorial representation of the overall flow — the three images below are from the real admin panel.

The problem

Handling requests consistently across three separate channels, routing them to the right department, and being able to respond outside working hours is a manageable operations problem, independent of the hospital's clinical quality.

Context

Özel Yenihayat Hastanesi receives patient requests across multiple channels, including Instagram, Facebook Messenger, and WhatsApp Business. When these channels run disconnected from each other, request tracking, appointment handling, and after-hours communication become fragmented.

The system

A unified inbox bringing conversations from Instagram and Facebook Messenger into one administrative interface; the hospital's AI communication assistant, Pınar; a structured appointment request workflow; an operator dashboard for hospital staff; developing operational reporting; ongoing work to connect WhatsApp Business; and work on an AI-supported voice reception layer for after-hours calls. Pınar is not a doctor, a diagnostic system, or an autonomous clinical decision-maker — human escalation and clear boundaries are a core part of the architecture, not an afterthought.

How one message moves through the system

Incoming message

Instagram or Facebook Messenger

Pınar assistance

Administrative questions, department routing — currently test accounts only

Appointment request

Collected in structured form

Human handoff

For anything complex or out of scope

Authorised hospital staff

Reviews and routes the request

What we delivered

  1. Unified communication infrastructure

    • Instagram Direct and Facebook Messenger connection, live via webhook
    • An administrative interface for managing conversations and handing them to staff
  2. Pınar — AI communication assistant

    • Answering administrative questions and recognising communication intent
    • Identifying the requested medical department and collecting appointment requests
    • Support outside working hours, with escalation to a human operator
  3. Appointment request workflow

    • Structured collection of appointment requests
    • Transfer of requests to authorised hospital personnel
  4. Operator dashboard and analytics

    • Conversation review, takeover, and approval workflows
    • Developing reporting on incoming requests and follow-up activity
Settings
The Relay Health admin panel's Settings screen: Meta/Instagram/Facebook shows 'Connected', Pınar AI is labelled 'Test accounts only', and WhatsApp and Phone/SIP show 'Not configured'.
From the real admin panel: the live connection status of every integration — what's working, what's in testing, and what hasn't been set up yet.
Analytics — leads
Relay Health analytics screen: 2 leads and 3 appointment requests over the last 30 days, broken down by channel, with untracked metrics explicitly marked as 'no data available'.
Real numbers from the last 30 days — small-scale but verifiable. Metrics with no real data source, like ad spend or revenue, are shown as 'no data available' rather than invented.
Analytics — performance
Relay Health analytics screen: a 1-minute average first response to appointment requests, and a 4.7-hour representative response time after handoff.
Operational performance: a 1-minute first response to appointment requests, with all 3 received fully processed.

Implementation

01

Human escalation and minimum necessary patient information were written into the architecture from day one — not a safeguard added later.

02

Pınar's scope was deliberately kept narrow: administrative questions, department routing, appointment requests. Clinical decision-making stays entirely out of scope.

03

Rather than replacing BizMed, the system was positioned to fill the communication gap around it.

Development status

Verified

  • Instagram and Facebook Messenger connection is live and working; the last 30 days brought 2 leads and 3 appointment requests through real channels.
  • Average first response to an appointment request is 1 minute; all 3 requests received were processed.
  • Pınar was tested against department-specific hospital communication scenarios.

In development

  • Pınar's automated replies currently run on test accounts only, ahead of a controlled rollout to real patient traffic.
  • The after-hours voice reception layer is in development and testing.
  • Operational reporting on incoming requests and follow-up activity is being developed.

Not yet

  • WhatsApp Business connection: trialled in development, not yet linked to a number on this hospital's live system.
  • Direct appointment booking through BizMed: the required API hasn't yet been provided by the hospital software vendor.

Technology

  • Instagram Direct and Facebook Messenger messaging integration (Meta)
  • AI-assisted conversation layer (Pınar)
  • WhatsApp Business API (trialled in development, not yet connected on this deployment)
  • Voice AI (in development)

Only verified scope appears on this page. No patient information or hospital operational data is shown.

Getting started

Where are you losing enquiries?

Tell us about the enquiry that goes unanswered, the follow-up done by hand, or the work that scatters between systems. In the first call we'll establish what can be systematised and where to start.

After you send the form

  1. A reply within two working daysWe read your message and suggest a time that suits you.
  2. A 30–45 minute discovery callA working session that maps the process, not a sales pitch. If it isn't a fit, we'll say so plainly.
  3. A written summaryAfterwards we put in writing what could be built, in what order, and within what scope.