Verity Health
Clinic website redesign with a voice AI receptionist and an SEO-safe migration.
Verity is a concept redesign for a clinic group with a site that had grown by accretion — years of pages added without structure, but with real search traffic that would be reckless to discard.
- AI pattern
- Voice AI agent
- AI component
- Voice AI receptionist
- Typical timeline
- ~8 weeks
- Maps to
- Website Premium + AI Agent
Verity Health is a fictional company — this is the build plan we execute for a business like it.
Try it live
Use it, don't just read about it.
A phone receptionist that books, reschedules and routes calls — live waveform, transcript, tool calls and hard safety guardrails. Take a call yourself by voice or text.
- 01Play the new-booking call (turn on voice to hear it)
- 02Try the urgent call: the agent stops and transfers
- 03Take a call yourself — speak if your browser supports it, or type
Scripted and deterministic so it runs anywhere. In production the same interface runs on live models, your data and your systems.
01The problem
What's really in the way.
The existing site was slow and dated but not worthless. Several deep pages carried genuine organic traffic, and a conventional redesign would have quietly deleted them.
- 01
Hundreds of URLs, many with inbound links, none of them inventoried.
- 02
Patients searching under stress, often on older devices, with accessibility as a hard requirement rather than a preference.
- 03
Sensitive content where tone and clarity matter more than visual ambition.
02What we'd build
The scope, plainly.
A redesign for a multi-clinic group that keeps its existing search traffic, plus a voice receptionist that answers the phone, books and reschedules against live availability, and routes anything clinical or urgent straight to a person.
Audit-led redesign
Every existing URL mapped and kept, merged or retired with a redirect before anything is cut.
Voice receptionist
Answers calls in natural speech, verifies the caller, and books, moves or cancels appointments against live availability — then texts a confirmation.
Safety guardrails
No diagnosis or medical advice. Urgent-sounding messages get the emergency guidance and the clinic phone number straight away.
Front-desk hand-off
Anything the assistant should not handle goes to reception with the conversation attached.
03Architecture
How the pieces fit.
Models are the swappable part. The tools, rules, logging and tests around them are what make a system dependable.
- PatientL1
- Redesigned website
- Phone line (voice agent)
- Web chat on the same brain
- Voice agentL2
- Streaming STT → LLM → TTS
- Urgency detection → emergency-number guidance (911 / 000) + transfer
- No-advice rules, identity check
- IntegrationsL3
- Twilio / SIP telephony
- Booking system API (availability)
- SMS confirmations, reception inbox
- OperationsL4
- Redirect map + 404 monitoring
- Conversation log, minimal data kept
- Next.js
- TypeScript
- Realtime voice (OpenAI Realtime / Vapi / Retell)
- Deepgram / Whisper STT
- ElevenLabs / OpenAI TTS
- Twilio
- Booking system API
- Vercel
04Screens
The interface, working.
Captured from the live demo in a headless browser — not mockups.
05Deliverables
What you'd own at the end.
- URL inventory and complete redirect map
- Redesigned templates at AA contrast
- Voice receptionist connected to live availability and SMS
- Guardrail rules reviewed with the clinic
- Reception hand-off inbox
- Post-launch search and 404 monitoring for the support window
06Typical timeline
Shipped in ~8 weeks.
- Audit & URL inventoryTraffic, links, keep list
- Design & templatesTask-based navigation, accessibility
- Build & redirectsStatic build, redirect layer, metadata parity
- Voice agent & guardrailsTelephony, availability, safety call tests
- Migration & monitoringLaunch, crawl checks, fixes
Schedule fixed in writing at kickoff.
07Approach
Decisions we'd take first.
Strategy
Audit first, then decide. Nothing was redesigned until we knew what it was doing.
- Inventory everythingEvery URL mapped against traffic, search position and inbound links before a single page was cut.
- Task-based structureNavigation rebuilt around what patients arrive trying to do, not the clinic's internal departments.
- Plan the migration firstThe redirect map was drafted alongside the sitemap, not discovered during launch week.
Design
Calm, high-contrast and unhurried. Large type, generous spacing and a palette that reads as clinical without feeling cold.
- Legibility as the briefBase type sized up, measure tightened, and contrast held well above the AA minimum throughout.
- WayfindingPersistent breadcrumbs and clear page-level orientation for people arriving deep from search.
- Plain languageClinical terminology paired with plain-language explanations rather than replaced by it.
Engineering
Rebuilt on static rendering with a complete redirect layer and structured data matching the visible content exactly.
- Redirect coverageEvery legacy URL resolves — mapped, consolidated or explicitly retired with a replacement.
- Metadata parityTitles, descriptions and canonicals carried across page by page, then verified after launch.
- Accessibility verifiedKeyboard and screen-reader paths tested on every template, not sampled.
08Intent
What it's designed to do.
Success criteria are agreed and measured from launch.
- Protect existing search traffic through the redesign.
- Make booking easier for patients on older phones and assistive tech.
- Keep the assistant inside clear safety limits, with a person one tap away.
- Answer every routine call at any hour without hiding the route to a person.
Your project
Want a plan like this for your business?
Book a 30-minute call. If it's a fit, we'll send a written scope and fixed price, or suggest starting with a $500 pilot.
Or email hello@xaevora.com · reply within one working day