Web and voice interface
Use a chat interaction or browser speech input, with browser voice output where supported.
New care-coordination range
Local product demonstratorMeridian Virtual Receptionist
Answer approved service questions, capture requests and prepare a clear handoff through a web or browser-voice reception experience.
An informed front door. A human-owned handoff.
The receptionist helps a visitor explain what they need and gives the team a structured request to review. It separates service information, appointment preferences and contact capture from the actual authority to book, verify identity or make a clinical decision.
What the AI does. What people decide.
The actual assistance
Service questions are answered using retrieved, approved public-facing Virtual Agent knowledge. The interface displays references. If the operator separately enables the configured language model, it can phrase an answer from those retrieved passages. Unmatched questions and requests for a person are directed to human review. Request capture and fixed emergency wording are bounded workflow behaviour, not autonomous clinical triage.
Human ownership
Appointment and arrival requests are not confirmed bookings or verified check-ins. This is a synthetic local reception demonstrator; unattended clinical triage and external service actions are not provided.
Features in the current workspace.
Reception request types
Capture appointment requests, arrival/check-in requests, referral enquiries and prescription enquiries for review.
Contact and preferences
Collect name, callback number, preferred appointment time, preferred contact route and a session summary.
Approved service answers
Look up opening hours and other service information from the permitted approved knowledge scope; display retrieved source passages.
Separate model opt-in
Allow configured model assistance explicitly for answers grounded in the retrieved approved content.
Human handoff
Request a person directly; route unsupported knowledge questions into the human-review path.
Fixed emergency response
Recognised emergency wording receives a fixed response and handoff boundary. The product does not determine clinical urgency.
Consented local save
Choose whether the simulated session may be saved. Review captured fields before adding the request to the local queue.
Reception activity
Inspect saved simulator-session counts, human handoffs, pending review and the explicit confirmed-bookings count.
Session reset
Start a fresh reception session without treating a previous visitor’s details as the next visitor’s context.
From first input to a reviewed next step.
- 01
Start a session
Choose web or supported browser voice interaction.
- 02
Understand the request
Capture the visitor’s wording and contact preferences.
- 03
Retrieve service information
Inspect approved references before presenting an answer.
- 04
Review the handoff
Confirm request details and whether a person is needed.
- 05
Save the request
Record a consented simulator request for human follow-up.
How the product works.
| Area | Mechanism and scope |
|---|---|
| Knowledge scope | Reception uses the approved, public-facing Virtual Agent content scope. A document merely existing in storage does not make it eligible. |
| Optional generation | Model-assisted answers use the configured Meridian provider after opt-in, with the retrieved passages as their answer context. |
| Local session storage | Saved simulator sessions enter the local application store and human-review queue. Counts describe that store, not a live contact centre. |
| Integration boundary | No live telephony, calendar availability, confirmed booking, outbound messaging or EPR connection is included. |
Product evidence and source notes
Content checked against the v25.2.0 source and visible local workspace on 30 October 2026. This is a product description and interface inspection, not an end-to-end deployment or clinical validation.
routes/ai_scribe.py, reception routes; static/new-care-products.js; RELEASE_NOTES_v25.2.0.md
Review assurance scopeQuestions about Virtual Receptionist.
Does it book an appointment?
It captures a request and preferred time. An authorised scheduling integration and explicit booking confirmation would be needed to make a real appointment.
Can it answer anything a visitor asks?
No. It uses approved service content and hands unsupported questions to a person rather than claiming an unrestricted answer.