Client Context
Venlyn OS targets US service businesses (e.g. plumbing, detailing) where missed calls and manual qualification directly reduce revenue. The product centralizes voice-led intake, CRM records, scheduling, and follow-ups in a compliant PWA.
The Challenge
Before
Owners and techs miss calls while on jobs; qualification questions repeat; scheduling is chatty and inconsistent.
Better
Leads need structured capture, service-radius validation, and auditable automation—not a fragile demo integration.
- Manual lead tracking and forgotten follow-ups
- Geographic waste: time spent on out-of-area callers
- Compliance and security cannot be bolted on late
The system had to be idempotent, observable, and resilient when AI or maps APIs degrade.
Solution Architecture
Next.js 15 + React 19 PWA with offline-oriented caching, push for critical alerts, and Zustand + TanStack Query for client state.
Supabase Postgres + RLS for multi-tenant CRM data; API routes for sensitive operations; rate limiting and hardening at the edge.
AI voice pipeline with webhooks, transcription, structured intent extraction, and a state machine for qualification/booking—synced to CRM via event-driven writes with deduplication.
Google Maps for geocode, distance/radius checks, and caching to control cost.
Key Implementation Decisions
Idempotent CRM writes from voice
Prevent duplicate clients/appointments; transactional creation with validation layers before commit.
Realtime-first UX
Subscriptions + optimistic updates so owners see calls and CRM changes without refresh.
Radius logic with graceful edge cases
Borderline addresses become follow-up tasks instead of hard rejections—preserves leads professionally.
Compliance by design
CCPA-oriented flows: access controls, deletion paths, audit trails for AI-touched records.
Operational Impact
- Fewer missed calls via always-on voice intake.
- Faster, consistent qualification without owner time on every ring.
- Clear follow-up queue for borderline or escalated cases.
- Owners operate from one CRM instead of spreadsheets plus voicemail.
Results
- Structured lead and appointment data from phone conversations.
- Improved response discipline through automated logging and follow-up tabs.
- Location-aware handling reduces wasted dispatches.
- Security and compliance patterns suitable for US production rollout.
Technology Stack
- Frontend
- TypeScript, React 19, Next.js 15, PWA (SW, manifest), TanStack Query, Zustand
- Backend
- Supabase (Auth, PostgreSQL, RLS), Next.js API routes
- AI & telephony
- Voice webhooks, NLP/intent pipeline, conversation state machine
- Maps
- Google Maps Platform (geocoding, distance/radius)
- Compliance
- CCPA-oriented controls, audit logging, secure secret handling