> ## 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 AI Calling? How AI Makes and Answers Phone Calls > AI calling uses voice agents to make and receive phone calls automatically — handling conversations in natural language without a human operator. How AI calling works, inbound vs outbound, latency, compliance, and how to build AI calling on Vobiz. *November 2026 · Vobiz Team* **AI calling is what happens when a phone call is handled entirely by an AI voice agent — no hold queue, no human operator, no "press 1 to continue."** The AI picks up, understands what the caller says, responds intelligently, takes action (booking, routing, logging), and either resolves the call or transfers to a human when needed. The same AI can run on the other side — dialing out to customers for reminders, outreach, or surveys — at a scale no human team can match. > [!NOTE] > > **Key takeaways** > > * AI calling = voice AI agent handling phone calls inbound and/or outbound > * Inbound: AI answers customer calls, routes by intent, resolves or escalates > * Outbound: AI dials contact lists, delivers messages, books appointments, qualifies leads > * Latency is 500–900ms — fast enough for natural conversation > * Vobiz handles the telephony (phone numbers, PSTN routing); AI platforms handle the conversation ## Two Types of AI Calling ### Inbound AI Calling A customer dials your phone number. An AI agent answers, greets them, understands their query in natural language, and either resolves it or routes it appropriately. ``` Customer dials your number ↓ Vobiz routes call to AI agent ↓ AI agent: "Hi, this is Priya from Acme Corp. How can I help?" ↓ Customer: "I want to know the status of my order" ↓ AI: [calls order API with customer's phone number] "Your order #4521 was shipped on 7th November and is expected to arrive today before 6pm." ↓ Customer hangs up. Call resolved — no agent involved. ``` **Inbound AI calling use cases:** * Customer support (order status, FAQs, billing) * Appointment scheduling and confirmation * Reception / call routing ("Which team do you need?") * After-hours handling * Lead capture from marketing campaigns ### Outbound AI Calling Your system initiates calls to a contact list. An AI agent runs the conversation when the contact answers. ``` System pulls contact from campaign list ↓ Vobiz dials the number ↓ AMD: human or machine? ↓ (human answers) AI agent: "Hi, this is calling from Acme. Your EMI of ₹4,500 is due on 15th. Press 1 to confirm, 2 to speak to our team." ↓ Contact presses 1 ↓ AI: "Thank you for confirming. Have a good day." Logs outcome to CRM. Dials next contact. ``` **Outbound AI calling use cases:** * Payment and EMI reminders * Appointment reminders and confirmations * Sales prospecting (AI cold calling) * Delivery status alerts * Survey and NPS collection * Reactivation campaigns ## How AI Calling Sounds Natural Three things determine how natural an AI call feels: **1. Latency**: The gap between when a person stops speaking and when the AI starts responding. Human conversational gap is 200–300ms. AI calling systems under 800ms total latency feel natural. Above 1,500ms, the pause becomes noticeable and awkward. **2. Voice quality**: TTS engines like ElevenLabs produce ultra-realistic voices that are often indistinguishable from human. For Indian callers, an Indian-accented voice (Sarvam AI, Google TTS Hindi) builds more trust than a foreign-accented AI voice. **3. Interruption handling**: Real conversations involve talking over each other. A good AI calling system detects when the caller interrupts, stops speaking, and listens — it doesn't barrel through its script ignoring what the caller said. ## AI Calling Technology Stack ``` Phone Network (PSTN) ↕ SIP Vobiz (telephony layer) ↕ SIP / WebSocket AI Agent Platform (Retell, Vapi, ElevenLabs, Custom) ├── ASR: Deepgram / Whisper / Sarvam ├── LLM: GPT-4o / Gemini / Claude └── TTS: ElevenLabs / OpenAI / Sarvam ↕ External APIs (CRM, calendar, order system) ``` **Vobiz role**: Provides phone numbers, SIP trunking, PSTN connection, call recording, webhooks, and India-specific features (140-series, NDNC filtering, GST invoicing). **AI platform role**: Runs the ASR → LLM → TTS pipeline, manages conversation state, executes function calls (API lookups, booking), and handles transfer/hangup actions. **Your server**: Provides business logic — which API to call, what data to look up, escalation rules, campaign contact lists. ## Setting Up AI Calling on Vobiz **For inbound AI calling:** 1. Buy a phone number on [Vobiz Console](https://console.vobiz.ai) 2. Create an inbound SIP trunk and set the origination URI to your AI platform's SIP address 3. Link your number to the trunk — calls now route to your AI agent 4. Configure the agent with your business logic, FAQs, and API integrations **For outbound AI calling:** 1. Buy a 140-series number (required for B2C outbound in India) 2. Create an outbound SIP trunk with digest auth credentials 3. Configure your AI agent platform with the outbound trunk credentials 4. Upload your contact list to Vobiz Campaign Manager or use the Make Call API 5. Enable AMD (answering machine detection) to skip unanswered calls Guides by platform: * [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) * [AI voice agent guide](https://www.vobiz.ai/docs/solutions/ai-voice-agent.md) ## AI Calling vs Robocalling Robocalling plays a pre-recorded message to everyone in a list — there's no intelligence, no interaction, no understanding of what the recipient says. It's widely banned for consumer calling in many jurisdictions. AI calling is fundamentally different: the AI understands and responds to what the caller actually says. It can handle objections, answer questions, change direction based on the conversation, and decide whether to escalate. This is why AI calling is legal under proper regulatory compliance — it's a genuine interactive conversation, not a broadcast. ## Compliance for Outbound AI Calling in India **140-series numbers**: All B2C outbound calls must come from a 140-series number registered on the TRAI DLT platform. Vobiz provisions 140-series numbers with self-serve setup. **DLT registration**: Your entity and the purpose of the calls must be registered. See the [DLT guide](https://www.vobiz.ai/docs/best-practices/dlt-registration.md). **NDNC filtering**: Scrub your contact list against the National Do Not Call registry before dialing. **Calling hours**: 9am–9pm only. No outbound commercial calls outside these hours. **Consent disclosure**: The AI agent must identify itself — "This is an automated call from \[Company]" — either at the start or when asked. For inbound AI calling (customers calling you), there are no outbound compliance requirements. Standard call handling rules apply. ## AI Calling Platforms Compared | Platform | Inbound | Outbound | Indian Languages | No-Code | | - | - | - | - | - | | **Retell AI** | ✅ | ✅ | Partial | ✅ | | **Vapi** | ✅ | ✅ | Partial | Partial | | **ElevenLabs** | ✅ | ✅ | Limited | ✅ | | **LiveKit** | ✅ | ✅ | Custom | ❌ | | **Sarvam AI** | ✅ | ✅ | ✅ Full | Partial | | **Custom (Python)** | ✅ | ✅ | Custom | ❌ | All connect to Vobiz via SIP trunk for the telephony layer. ## Next Steps * [AI outbound calling guide](https://www.vobiz.ai/docs/blogs/what-is-ai-outbound-calling.md) — campaigns, AMD, concurrency * [AI cold calling guide](https://www.vobiz.ai/docs/blogs/what-is-ai-cold-calling.md) — sales prospecting automation * [AI receptionist guide](https://www.vobiz.ai/docs/blogs/what-is-an-ai-receptionist.md) — inbound call handling * [Get a Vobiz number](https://console.vobiz.ai) and try your first AI call This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.