Palmview Dental — AI Customer Support Experience
An interactive AI customer-support concept demonstrating how businesses can provide consistent answers, organise enquiries and improve customer communication.
- Client
- Palmview Dental (fictional)
- Sector
- Healthcare
- Setting
- Abu Dhabi
- Status
- Concept project · not a client
Palmview Dental, its staff, address, hours and policies are invented for this concept. There are no real patients, results or testimonials.
Concept project. Runs entirely in your browser: a deterministic knowledge engine, not a live AI model, and nothing is sent.
Overview
Two products, one source of truth.
A calm website for a fictional general dental clinic in Abu Dhabi, with an assistant that answers visitors' questions, and a separate support app where staff manage what the assistant knows and follow up on what it couldn't answer.
Both read the same editable knowledge base and clinic details, so changing the opening hours in the app updates the website and the assistant together.
The problem
Why this matters.
Clinic front desks answer the same questions about hours, location, treatments and appointments all day, while after-hours enquiries wait until morning. Answers drift between the website, the phone and staff, and questions that need a professional can get a casual reply.
The solution
Consistent answers, honest limits.
One source of truth
The website and the assistant read the same editable knowledge base and clinic details, so an update to the hours changes both at once.
Grounded, scored answers
Each reply is matched to approved entries with a confidence score and a visible source. Low confidence means a clarifying question or a person, never a guess.
Safe hand-off and requests
Clinical or urgent questions are referred to professionals. Appointment requests and staff enquiries are collected with consent and land in a staff inbox.
Preview
Inside the demo.
Screens from the working demo, captured from the browser. Data shown is sample data.
Floating website assistant
Answers come from approved entries, show a confidence label and list their sources. A two-part question gets both parts answered.
Knowledge base manager
Staff create, edit, disable and test entries. A change applies to the very next reply, on the website and in the app.
Conversation history
Every exchange is kept with its detected intent, confidence against the threshold, and the entries it used.
Analytics from real local data
Counts are calculated from the conversations stored in the browser. Sample visitors are labelled and can be excluded.
How it answers
A pipeline you can inspect.
- 01
Normalise
Lower-case, expand contractions, map phrases like “check-up” and “how much” to known terms.
- 02
Classify
Detect the intent first, so urgent or clinical questions are routed to safety before anything else.
- 03
Retrieve & rank
Score every enabled entry with weighted term coverage, then rank by confidence.
- 04
Decide
Above the threshold: answer with sources. Just below: answer cautiously. Otherwise clarify or offer a person.
Features
- Clinic website with floating assistant
- Dedicated support app
- Editable knowledge base
- Confidence scoring & source attribution
- Appointment request flow
- Human handoff queue
- Unanswered-question tracking
- Analytics from real interactions
Technology
- Next.js
- React
- TypeScript
- Tailwind CSS
- Motion
- Local knowledge engine
- Browser storage
Designed to deliver
- Consistent answers across website and chat
- Enquiries captured after hours instead of lost
- Less repetitive work for reception
Design goals, not measured results.
What it is, and isn't
- It is a deterministic knowledge engine running in the browser, not a live AI model. It only says what the knowledge base says.
- Nothing is sent anywhere. Requests, conversations and settings are stored in the visitor's own browser.
- An appointment request is recorded as a demonstration and is never a real booking.
- The engine sits behind a provider interface, so a hosted model could be added later without changing the interface.
Try it yourself
Ask the assistant about hours, parking or treatments.
Next demo
Harbor Key Realty — Real Estate CRM & Workflow Automation
Want an assistant for your business?
We can build one around your own services, hours and policies. Tell us how enquiries reach you today and we'll reply with a scope and fixed quote.