Review every AI voice call against your own rules.
GoldNovaAI checks the calls your voice agents handle against a policy you control, and shows the exact transcript quote behind every flag. Your team reviews evidence, not hunches.
Works with any voice stack through our API. Results are QA signals for human review — not legal or compliance determinations.
We guarantee a full refund.
Hello, this call is recorded for quality purposes.
From call to evidence-backed finding
The part of voice-agent QA that usually means listening to recordings and filling in a spreadsheet.
Send your calls
Post transcripts through the API, upload recordings in the dashboard, or connect Vapi, Retell or Twilio (developer preview).
Apply your policy
Start from a default policy and edit it: required phrases, prohibited claims, sensitive data, escalation, script steps.
Get findings with evidence
Every potential violation carries a verbatim quote from the transcript. The database refuses to store one without it.
Review and decide
Open the call report, jump to the moment in the audio, and decide. The tool recommends; people decide.
What it checks
Eight rule types you can switch on, edit and tune. Rules match plain text you write — no regular expressions to maintain.
Built to be checked, not trusted blindly
Evidence or no violation
A potential violation must quote the transcript, and the quote must appear in the call. Both are enforced in the database, not just in the app.
Careful wording
“No issues detected”, “Review recommended”, “Potential violation”. Findings are signals for a human reviewer — never a legal conclusion.
AI is optional and constrained
Rules are deterministic by default. Optional AI assessment can raise a finding, but it cannot clear a violation, and its quotes are verified against the transcript.
No guessing who spoke
Speech-to-text has no speaker labels. When the speaker matters and is unknown, the result is “Review recommended” instead of a guess.
Your data stays separate
Each organization is isolated in the database. Recordings are private and played through short-lived signed links. We do not train models on your calls.
An API from day one
API keys, idempotent ingestion, and a documented result format, so QA fits into the stack you already run.
Connects to your voice stack
| Source | Status | Notes |
|---|---|---|
| Generic API (transcripts) | Available | Idempotent, API-key authenticated. |
| Recording upload | Available | Speech-to-text, speakers not labelled. |
| Vapi · Retell · Twilio | Developer preview | Signed webhooks, tested against documented payloads; not yet tested with live provider accounts. |
Using another platform? Send transcripts through the API, or tell us what you need at hello@goldnovaai.com.
More from GoldNovaAI
WhatsApp SaaS Infrastructure
Developer previewA multi-tenant foundation for products that talk to customers on WhatsApp. Not generally available yet.
Research Evidence Workspace
AvailableTurn interviews into evidence-linked quotes, themes and findings, with uncertain transcript sections flagged for review.
Try it on your own calls
Voice QA is in early access and free to evaluate.