This documentation page is live-rendered from the AI WCRM Knowledge Engine.
# Multi-Provider Voice AI Setup (ElevenLabs & Retell AI)
AI WCRM features a provider-agnostic **Multi-Provider Voice AI Architecture**. You can deploy human-grade AI voice agents using **ElevenLabs Conversational AI** or **Retell AI**, with native support for Indian languages (Hindi, Hinglish) and automatic post-call CRM intelligence extraction.
---
### Core Platform Capabilities
<Callout type="tip">
**Provider Agnostic**: AI WCRM owns call orchestration, CRM intelligence extraction, and cost governance. Retell AI and ElevenLabs act as interchangeable voice adapters — switch anytime without losing CRM data!
</Callout>
- **Multi-Provider Support**: Choose between Retell AI (~620ms latency) and ElevenLabs (~310ms latency).
- **Native Hindi Voice Agents**: Deploy ElevenLabs native Hindi voices (**Priya** & **Arjun**) for localized student and customer outreach.
- **Shared Knowledge Base**: Voice AI agents automatically consume the exact same Knowledge Base, AI Rules, Personality, and Business Hours configured in your AI Assistant settings.
- **Cost Governance**: Set monthly INR budgets, daily call caps, and per-user call limits to prevent runaway spending.
- **10-Field Post-Call Intelligence**: Every completed call is analyzed by your BYOK LLM (Gemini, OpenAI, Groq) to extract structured CRM fields.
---
### Step 1: Configure Provider Credentials in AI WCRM
1. Navigate to **AI Assistant → AI Voice Calling** (or **Settings → AI Voice**).
2. Choose your active provider tab (**Retell AI** or **ElevenLabs**).
3. Input your provider API Key and Agent ID:
- **Retell AI**: Obtain API Key and Agent ID from [retellai.com/dashboard](https://retellai.com/dashboard).
- **ElevenLabs**: Get API Key and Conversational Agent ID from [elevenlabs.io/app](https://elevenlabs.io). For Hindi outreach, select **Priya** or **Arjun** voice models.
4. Click **Save Configuration** and run **Test Connection** to verify API health.
---
### Step 2: Set Up Cost Governance Guardrails
Prevent billing surprises by configuring spending limits in **Settings → AI Voice → Cost & Limits**:
- **Monthly Account Budget**: e.g., ₹5,000 / month. Calls are automatically paused if budget is reached.
- **Daily Call Cap**: Maximum outbound/inbound calls per day.
- **Per-User Limit**: Maximum daily calls an individual sales agent can trigger.
---
### Step 3: Post-Call Intelligence & Ecosystem Sync
When a voice call completes, AI WCRM automatically extracts 10 structured fields and syncs across 9 CRM modules:
```ts
// 10-Field Post-Call Structured Intelligence
export interface VoiceCallAnalysis {
summary: string; // 2-3 sentence executive summary
customerIntent: string; // Main customer request/question
sentiment: 'positive' | 'neutral' | 'negative';
buyingSignals: string[]; // e.g. ["asked about discount", "requested campus tour"]
objections: string[]; // e.g. ["budget constraint"]
nextFollowupAt?: Date; // ISO timestamp for follow-up
actionItems: string[]; // e.g. ["Send prospectus on WhatsApp"]
aiLeadScore: number; // 0–100 calculated score
opportunityStage: 'cold' | 'warm' | 'hot' | 'closed';
aiRecommendation: string; // Next best action for sales team
}
```
#### Automatic Module Sync Map:
1. **Contacts**: Updates Lead Score, Customer Intent, Sentiment, and Last Called timestamp.
2. **Deals Pipeline**: Advances deal stage (e.g. Warm → Hot) and flags priority deals.
3. **Tasks**: Auto-creates follow-up tasks with due dates and recommended actions.
4. **WhatsApp**: Automatically sends requested documents (brochures, fee charts) via WhatsApp.
5. **Decision Center**: Pushes high-priority AI recommendation cards for management review.
---
<Callout type="info">
**Fallback Mechanism**: If the primary voice provider experiences downtime, AI WCRM automatically routes outbound calls to the configured backup provider.
</Callout>