VoxSerene — Call intelligence
Call intelligence that holds up to a UK data review
Every call captured, transcribed, and analysed per location — with consent, retention, and access handled as product requirements. Multi-tenant, role-scoped, and built so a data-protection question has an answer in the system rather than in a spreadsheet.
The problem
UK operators face the same blind spot as anyone else — the phone is the busiest channel and the least measured — but they can't solve it by pointing a US analytics tool at their call recordings. Recordings are personal data. Retention windows, lawful basis, subject access, and audit expectations are not features you enable after launch; a platform that treats them that way becomes the compliance problem it was meant to remove.
What changes
The same operational clarity — transcripts, sentiment, topics, per-location reporting, routed follow-up — on a footing that survives scrutiny. Retention is explicit, access is scoped by role and enforced in the database, and what the platform did with a record is in an audit trail rather than in someone's memory.
Built for
What it does
Inside the platform
Multi-source call ingestion
Signature-verified webhook ingestion from voice-agent and telephony sources, normalised into one call record regardless of where the call came from.
Transcription and analysis
AI transcription with sentiment, topics, and action items attached to each call, with re-enrichment available when an upstream job fails.
Retention and audit
Explicit retention handling over calls, recordings, and audit logs — so what is kept, for how long, and who looked at it are properties of the platform.
Per-location reporting
Dashboards, search, and trend reporting scoped to the sites a user is responsible for, with exports for the reporting that happens outside the product.
Multi-site tenancy
Tenant → location → phone-number structure with per-tenant branding and onboarding, so a group runs its sites separately under one account.
Routed notifications
Branded email notification with per-location recipients, keeping call follow-up with the team that owns the site.
How it's built
The stack, specifically
No mystery boxes. These are the actual choices behind the platform — the ones that decide how it scales, who can read what, and how quickly it can change.
Frontend
- React + TypeScript + Vite
- Tailwind CSS + Radix UI
- Role, permission, and tenant-branding contexts
Backend
- Supabase Edge Functions (Deno/TypeScript)
- Ingestion and enrichment workers
- APIs for calls, reporting, search, notifications, and administration
Data & governance
- Postgres with row-level security per tenant and location
- Supabase Storage for recordings and transcripts
- Audit logging with defined retention
- Migration-only schema evolution
Our role
Built and operated by Narix Labs
The ingestion and enrichment pipeline, the multi-tenant data model and its security policies, the admin application, and the governance surface — audit logging, retention, and the test suites that prove access rules hold at the database level.
Isolation enforced in Postgres
Tenant and location boundaries are row-level security policies with their own SQL tests, rather than filters the frontend is trusted to apply.
Governance as a feature surface
Audit logs and retention are built as product capabilities with their own screens and policies — the parts of a call platform that usually get deferred until an audit forces them.
FAQ
VoxSerene questions
Need something like VoxSerene?
Tell us what you're holding — an idea or a codebase — and you'll have a written proposal for the first step within 48 hours.