Every feature, in plain language.
AI evaluation is the engine. Around it sits everything your team needs to run a serious QA program — and a few things they didn't know they could ask for.
Every conversation, scored against your rubric.
Belter applies your custom criteria — empathy, problem resolution, compliance, product knowledge, anything you score on — to 100% of conversations. Weighted, versioned, and audit-friendly.
- Customizable criteria with per-criterion weighting
- Versioned prompts — rubric changes don't break old scores
- JSON-structured output with reasoning per criterion
- Bring-your-own OpenAI key or use ours
- Agent notifications + webhooks on completion
Connect anything. Score everything.
Native connectors for the helpdesks and call platforms you already use, plus a generic ingestion API for everything else. Scheduled sync from 15-minute to daily, with "all missing" or "since last fetch" modes.
- LiveChat, Chatwoot, Zendesk, Intercom, Upmind
- RingCentral phone with Whisper transcription + diarization
- Amazon Connect via Contact Lens webhook
- Generic REST API for Genesys, NICE, Five9, Talkdesk, ServiceNow
- Circuit breaker on integration failure
Where AI scoring meets human judgment.
AI handles coverage. Your QA team handles consistency. Audit sessions, blind calibration, and assignment workflows that make the whole program audit-ready.
- Audit Sessions — up to 10 per agent / week, AI-suggested
- Calibration Sessions — multiple analysts score blind
- FCR, authentication, contact driver, issue/trend capture
- QA assignments — analysts mapped to specific agents
- Analyst-specific QA dashboard
3 analysts · 5 conversations · Blind
| Analyst | C1 | C2 | C3 | C4 | C5 |
|---|---|---|---|---|---|
| Harry B. | 87 | 94 | 71 | 88 | 92 |
| Aisha K. | 85 | 91 | 74 | 86 | 90 |
| Bruno S. | 88 | 93 | 58 | 87 | 91 |
| Variance | ±1.5 | ±1.5 | ±8.0 | ±1.0 | ±1.0 |
C3 needs discussion — variance above threshold. Close session when aligned.
From scorecard to coaching note in three clicks.
The coaching wizard picks an agent, surfaces eligible conversations (date filters, "already coached" badges), and lets Belter draft the notes for you. Coaching links back to the audit. Agents get notified. Nothing falls through.
- 3-step coaching wizard tied to real conversations
- AI-generated coaching notes from evaluation results
- Goals tied to QA metrics — set, track, automate review
- Agent self-assessment against the same rubric
- Gamification + leaderboards + badges (opt-in)
- Product Knowledge dipped 8pt on billing conversations this week.
- Strong empathy patterns — call out the Maria Rodriguez recovery.
- Suggest the "Credit Policy" training module (LMS).
Catch problems before the regulator does.
A rule engine for keywords, phrases, regex and required disclosures — paired with built-in PII detection and semantic checking. Risk levels surface in real time.
- PII detection: email, credit card, SSN, UK NI, IBAN, phone
- Semantic disclosure checking via LLM (paraphrases pass)
- Risk levels: low / medium / high / critical
- Smart alerts + manager notifications on high/critical
- Custom keyword / phrase / regex / disclosure rules
Daily anomaly detection. No-code workflows.
Belter watches your scores, your CSAT, your volume, and individual criteria — then sends the right alert to the right person at the right threshold. Build downstream workflows with a trigger/condition/action builder.
- Daily anomaly detection across all metrics
- Per-org thresholds, per-user notification preferences
- Inbox with acknowledge workflow
- Automation builder: trigger → condition (all/any) → action
- Triggers: evaluated, coaching created, CSAT captured, etc.
Know why customers contact you.
AI classifies every conversation by primary and secondary contact driver. Embeddings (pgvector) match against your existing driver taxonomy — and grow it automatically as new themes emerge.
- Primary + secondary driver classification
- Auto-growing taxonomy via vector similarity
- Driver volume × quality correlation
- Repeat-contact root-cause analysis
For holding companies. For agencies. For anyone with more than one brand.
Belter is built around the assumption that your org runs multiple brands. Auto-derive from integration source, override at import, merge duplicates, and report at the brand level.
- Auto-derive brand from integration data
- Manual override at import time
- Brand aliases, merge, CRUD
- Brand-level performance reporting
Enterprise-ready from day one.
Granular RBAC, SSO, audit logs, scoped API tokens, and per-org Stripe billing. Belter ships with everything procurement and security ask for in week one — not as a six-month upsell.
- 8 default roles + custom; team and self scoping
- SSO / SAML with domain routing and JIT provisioning
- 2FA (TOTP + recovery codes); enforceable per-org
- Audit log for sensitive actions
- API tokens with scoped permissions; Stripe billing
- Agent impersonation for managers (with audit trail)
Ready to see it on your conversations?
100 conversations free. No credit card. Cancel any time.