> ## Documentation Index > Fetch the complete documentation index at: https://www.vobiz.ai/docs/llms.txt > Use this file to discover all available pages before exploring further. # What Is an AI Receptionist? How It Works, Use Cases & Setup > An AI receptionist is a voice agent that answers calls, routes callers, books appointments, and handles FAQs 24/7 — no hold music, no missed calls. How it works, when to use one, and how to build one on Vobiz. *November 2026 · Vobiz Team* **An AI receptionist is a voice agent that picks up every call, says the right thing, routes the caller, books appointments, and handles common questions — all in real time, without hold music or missed calls.** Unlike a human receptionist who works 8 hours a day and handles one call at a time, an AI receptionist works 24/7 and can run hundreds of simultaneous conversations. > [!NOTE] > > **Key takeaways** > > * An AI receptionist handles inbound calls end-to-end: greeting, routing, FAQs, and appointment booking > * It runs on three layers: telephony (the phone connection), ASR+LLM (the AI brain), and TTS (the voice) > * Vobiz provides the telephony layer — connect any AI agent platform via SIP trunk or WebSocket > * Cost is typically ₹1–3/minute all-in, versus ₹25–40/minute for a human receptionist's time > * Indian language support (Hindi, Tamil, Telugu, Marathi and more) is available via platforms like Sarvam AI ## What Does an AI Receptionist Do? An AI receptionist handles the calls that would otherwise go to a front-desk employee or ring out unanswered. The core capabilities: | Capability | What it means | | - | - | | **Answer every call** | No missed calls, no voicemail during business hours or after | | **Greet callers by name** | Pull caller name from CRM using their phone number before they say a word | | **Route to the right person** | "Press 1 for sales" or voice-based: "I'll connect you to billing" | | **Answer FAQs** | Hours, location, pricing, appointment status — no agent needed | | **Book appointments** | Check calendar availability and confirm booking live on the call | | **Take messages** | Log caller intent, name, and callback number, then notify your team | | **Escalate gracefully** | Transfer to a human agent when the caller asks or the issue is complex | | **Speak Indian languages** | Hindi, Tamil, Telugu, Marathi, Bengali — route by language preference | ## How an AI Receptionist Works Every AI receptionist sits on three layers: ``` Caller → Phone Network → Telephony Provider (Vobiz) ↓ SIP Trunk / WebSocket ↓ AI Voice Agent Platform (ASR → LLM → TTS → Response) ↓ Back to caller in ~700ms ``` **Layer 1: Telephony (Vobiz)** — Your business phone number is owned by a telephony provider. When a call comes in, the provider routes it to your AI agent over SIP or WebSocket. Vobiz handles the PSTN bridging, call recording, concurrent channels, and India-specific numbers (10-digit mobile, 140-series business). **Layer 2: ASR + LLM** — Automatic speech recognition (Deepgram, Whisper, Google STT) transcribes the caller's words in real time. A large language model (GPT-4o, Gemini, Claude) reads the transcript, decides what to do, and generates a response — look up availability, transfer, or simply reply. **Layer 3: TTS** — Text-to-speech (ElevenLabs, OpenAI TTS, Sarvam) converts the agent's reply to natural-sounding audio and plays it back to the caller. The full loop — speech in, text transcribed, LLM responds, TTS plays back — completes in 500–900ms, fast enough to sound like a real conversation. ## AI Receptionist vs Human Receptionist | Factor | Human Receptionist | AI Receptionist | | - | - | - | | **Availability** | 8–9 hrs/day, weekdays | 24/7, 365 days | | **Simultaneous calls** | 1 | Unlimited | | **Cost (India)** | ₹25,000–40,000/month | ₹1,000–8,000/month (depends on volume) | | **Languages** | 1–2 fluently | 20+ with multilingual platforms | | **Consistency** | Varies by mood, fatigue | Always consistent | | **Ramp time** | Weeks of training | Hours to configure | | **Complex issues** | Handles well | Escalates to human | | **Appointment booking** | Manual calendar check | Real-time API integration | The AI receptionist is not a replacement for every human interaction — it's a first-layer filter that resolves the 60–80% of calls that don't need a human, so your team handles only the ones that do. ## Use Cases by Industry **Healthcare clinics and hospitals** Patients call to book appointments, ask about doctor availability, or get directions. The AI receptionist handles all three without queue wait. It connects to the scheduling system, reads available slots, and books on the spot. For prescription refill or test results, it transfers to the clinical team. **Legal and professional services** A law firm's AI receptionist answers after-hours calls, collects the caller's name and matter type, and schedules a callback or consultation. No potential client hears voicemail. **Real estate agencies** Callers enquiring about a listed property get the AI agent, which reads the property details, schedule a site visit, or transfers to the responsible agent. Runs all weekend without an on-call person. **Restaurants and hospitality** Table bookings, delivery status, menu questions — all handled by voice. No staff member answers the phone during the dinner rush. **E-commerce and retail** Order status, return requests, delivery queries — most callers just want their order status. The AI reads it from your system and confirms verbally in under 30 seconds. **Small businesses and startups** A solo founder or small team can project a professional image by running an AI receptionist on their business number. All calls are answered, triaged, and routed appropriately without a full-time hire. ## Building an AI Receptionist on Vobiz Vobiz provides the phone number and SIP/WebSocket connection. You choose the AI agent platform. The two connect via SIP trunking (for platforms like Retell AI, Vapi, ElevenLabs) or WebSocket streaming (for custom-built agents). **Step 1: Get a phone number** Buy a Vobiz number from the [console](https://console.vobiz.ai) — Indian 10-digit mobile, 140-series business line, or international DID. This is your receptionist's public number. **Step 2: Choose an AI agent platform** | Platform | Best for | | - | - | | **Retell AI** | No-code agent builder with built-in call flows | | **Vapi** | Developer-first, function calling, custom LLM support | | **ElevenLabs** | Ultra-realistic voice quality | | **Sarvam AI** | Hindi and Indian language support | | **Custom (Python/Node)** | Full control over ASR, LLM, TTS pipeline | **Step 3: Connect via SIP trunk** Create an outbound SIP trunk on Vobiz, copy the domain and credentials, and paste them into your AI agent platform. For inbound, create an inbound trunk pointing to the platform's SIP URI. Detailed guides: * [Retell AI + Vobiz](https://www.vobiz.ai/docs/integrations/retellai-dashboard.md) * [Vapi + Vobiz](https://www.vobiz.ai/docs/integrations/vapi-dashboard.md) * [ElevenLabs + Vobiz](https://www.vobiz.ai/docs/integrations/elevenlabs-dashboard.md) **Step 4: Configure the agent** Write a system prompt that describes your business, the agent's role, and what it should do in each scenario. Include: * Business name, hours, and location * FAQs and their answers * Escalation triggers ("if the caller says 'emergency' or 'urgent', transfer to +91XXXXXXXXXX") * Appointment booking instructions and calendar API credentials **Step 5: Test and go live** Call your Vobiz number. The AI should answer. Run through the main scenarios — booking, FAQ, transfer — and adjust the prompt until the behaviour matches your requirements. ## Latency and Voice Quality For a natural-sounding AI receptionist, end-to-end latency (from when the caller stops speaking to when the agent starts) should be under 1,000ms. Vobiz's infrastructure is co-located in Mumbai, reducing round-trip time for India-originated calls. | Component | Typical latency | | - | - | | ASR (speech-to-text) | 100–300ms | | LLM inference | 200–500ms | | TTS (text-to-speech) | 100–200ms | | Network (India) | 50–150ms | | **Total** | **450–1,150ms** | Choosing a low-latency TTS (ElevenLabs Flash, OpenAI TTS) and running LLM inference on a fast model (GPT-4o-mini, Gemini Flash) keeps the full loop under 800ms in most configurations. ## India-Specific Considerations **TRAI compliance**: Outbound calls from an AI receptionist fall under telemarketing regulations. Use a 140-series number for business communication, register on DLT if sending follow-up SMS, and honour NDNC/DND preferences. Inbound calls (where the customer is calling you) have no outbound compliance requirement. **Indian language support**: Hindi is widely supported across all platforms. For Tamil, Telugu, Marathi, Bengali, Kannada, and Gujarati, Sarvam AI provides high-quality regional ASR and TTS. See the [Sarvam integration guide](https://www.vobiz.ai/docs/integrations/sarvam.md). **Caller ID**: Configure your 140-series number as the displayed caller ID for outbound transfers and callbacks to reinforce brand trust. ## Next Steps * [Buy a phone number on Vobiz](https://console.vobiz.ai) and pick up your first Indian business number * Connect a Retell AI or Vapi agent using the [dashboard guides](https://www.vobiz.ai/docs/integrations/retellai-dashboard.md) * Read the [AI voice agent guide](https://www.vobiz.ai/docs/solutions/ai-voice-agent.md) for a full architecture walkthrough * Explore [Sarvam AI integration](https://www.vobiz.ai/docs/integrations/sarvam.md) for Indian language support This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.