> ## 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 IVR? Conversational IVR vs Traditional Menu Systems > AI IVR replaces rigid press-1-press-2 menus with a conversational AI agent that understands natural speech, routes intelligently, and resolves queries without a human. How AI IVR works, how it differs from traditional IVR, and how to build one on Vobiz. *November 2026 · Vobiz Team* **AI IVR is what interactive voice response becomes when you replace the menus with a conversational AI.** Instead of forcing callers through "Press 1 for Sales, Press 2 for Support, Press 3 for Billing, Press 4 to hear these options again", the AI simply asks "How can I help?" and understands whatever the caller says. The result: faster call resolution, dramatically lower abandonment, and an experience that feels like talking to a knowledgeable assistant rather than navigating a phone tree. > [!NOTE] > > **Key takeaways** > > * AI IVR = traditional IVR menus replaced with a conversational AI voice agent > * Callers speak naturally instead of pressing keys; AI understands intent and routes accordingly > * Reduces IVR abandonment by 40–60% vs traditional menus (no more callers pressing 0 to escape) > * 60–80% of calls can be fully resolved by AI IVR without human involvement > * Built on Vobiz telephony + an AI agent platform (Retell, Vapi, custom) connected via SIP ## Traditional IVR vs AI IVR | Factor | Traditional IVR | AI IVR | | - | - | - | | **Caller input** | DTMF keypad (1, 2, 3…) | Natural speech | | **Menu structure** | Fixed tree, 2–4 levels | No menus — open conversation | | **Unexpected input** | "I didn't understand that. Press 1 for…" | "Could you tell me more about that?" | | **Language support** | Pre-recorded audio per language | Live multilingual AI | | **Updates** | Re-record audio files | Edit a text prompt | | **Resolution rate** | Routes to agent (rarely resolves) | Resolves 60–80% autonomously | | **Caller experience** | Frustrating for complex needs | Natural and conversational | | **After-hours** | Voicemail or overflow | AI continues 24/7 | The traditional IVR is not dead — it still has a place for ultra-high-volume, simple flows (OTP delivery, payment confirmation, single-step DTMF capture) where cost per call matters more than experience. But for anything involving routing, information lookup, or multi-step interaction, AI IVR outperforms in every metric. ## How AI IVR Works ``` Caller dials your number ↓ Vobiz answers the call ↓ Routes to AI agent via SIP trunk or WebSocket ↓ AI agent: "Thank you for calling Acme. How can I help you today?" ↓ Caller: "I want to know when my delivery will arrive" ↓ ASR transcribes: "I want to know when my delivery will arrive" LLM classifies intent: ORDER_STATUS LLM calls order API with caller's phone number API returns: "Delivery scheduled for today, 3–6pm" TTS speaks: "Your delivery is scheduled for today between 3 and 6pm. Is there anything else I can help you with?" ↓ Caller: "No, thank you" → Call ends Call resolved — 0 agent minutes used ``` The entire interaction — ASR, LLM reasoning, API call, TTS — completes in under 1 second per turn. ## AI IVR Use Cases **Customer support routing**: "I have a billing question" → routes to billing. "My order hasn't arrived" → triggers order lookup. "I need to speak to a manager" → escalates immediately. No menu, no guessing. **Appointment booking**: "I'd like to schedule an appointment for Friday afternoon" → AI checks calendar API, offers available slots, confirms booking. No agent required. **Order and delivery status**: Caller's phone number is used to look up their most recent order. AI reads the status. Done in 20 seconds without queue wait. **FAQ resolution**: "What are your business hours?" / "Where are you located?" / "What is your return policy?" — all answered from the knowledge base without connecting to an agent. **Banking and fintech**: Account balance, last transaction, EMI due date — AI reads live from the core banking system using the caller's registered mobile number for identity. **Healthcare**: Appointment reminders and rescheduling, prescription status, lab report availability — the AI acts as the clinic's first point of contact, triaging and routing to the right department. **Multilingual routing**: Caller speaks Hindi → AI responds in Hindi. Caller says "Tamil lo matladandi" → AI switches to Tamil. No pre-recorded prompts, no "Press 2 for Hindi" menus. ## Building AI IVR on Vobiz **Step 1: Buy a phone number** Get an inbound number from [Vobiz Console](https://console.vobiz.ai) — 10-digit Indian mobile, 1800 toll-free, or 140-series business number. This is the number callers dial. **Step 2: Create an AI agent** Choose your AI platform and write the system prompt. For AI IVR, the prompt should define: ``` You are the automated customer service agent for [Company]. When a customer calls, greet them and ask how you can help. You can: 1. Look up order status (call /api/order-status with phone number) 2. Book appointments (call /api/book with date and time) 3. Answer FAQs: - Hours: Monday–Saturday, 9am–6pm - Location: [Address] - Return policy: 30 days, unused condition 4. Transfer to a human agent if the issue is complex Transfer if: - Caller is upset or uses angry language - Query involves refunds > ₹1,000 - Caller says "I want to speak to a human" Speak in [language]. Be brief and helpful. ``` **Step 3: Connect via SIP trunk** Create an inbound trunk on Vobiz, set the origination URI to your AI platform's SIP address, and link your number. All inbound calls now route to the AI. **Step 4: Configure function calling / API access** Connect the agent to your data sources: * Order management system: returns order status by phone number * Calendar API: checks availability, creates bookings * CRM: retrieves customer account data * Knowledge base: FAQ answers for common questions **Step 5: Set up escalation** When the AI decides to transfer, it uses a Dial action to connect to your human agent queue. Configure call whisper — a 5-second summary the agent hears before the customer bridges: "Order status query. Customer says package hasn't arrived. Order #4521, dispatched 5 days ago." Guides: * [Retell AI integration](https://www.vobiz.ai/docs/integrations/retellai-dashboard.md) * [Vapi integration](https://www.vobiz.ai/docs/integrations/vapi-dashboard.md) * [AI voice agent guide](https://www.vobiz.ai/docs/solutions/ai-voice-agent.md) * [Cloud IVR guide](https://www.vobiz.ai/docs/solutions/cloud-ivr.md) ## Hybrid: AI + Traditional IVR You don't have to choose one or the other. A hybrid approach is common: ``` Call arrives → Traditional IVR (single-level) "Press 1 for Sales, 2 for Support, 3 for Billing" ↓ Caller presses 2 (Support) ↓ AI IVR handles support queries: "What's your support issue today?" ↓ AI resolves or routes to human support agent ``` The first-level DTMF menu reduces LLM cost by pre-classifying call intent — the AI only needs to handle the conversation within a narrower domain. This is useful for businesses with very distinct departments where callers know exactly which department they need. ## AI IVR Cost | Component | Cost | | - | - | | Vobiz phone number | ₹200–500/month | | Vobiz inbound minutes | ₹0.20–0.45/min | | AI platform | ₹2,000–8,000/month or per-min | | LLM + TTS inference | ₹0.50–2.00/min of conversation | | **Per resolved call (avg 2 min)** | **₹1.00–5.00** | Compare: an agent-handled call in an Indian contact center costs ₹30–80 fully loaded. At 500 calls/day, AI IVR resolving 70% autonomously saves ₹10,500–28,000/day vs routing all calls to agents. ## Measuring AI IVR Performance | Metric | Description | AI IVR Target | | - | - | - | | **Self-service rate** | % of calls resolved without agent | > 65% | | **IVR abandonment rate** | % who hang up during AI interaction | \< 10% | | **Intent recognition accuracy** | % of callers correctly understood | > 90% | | **Average resolution time** | Time from answer to call end | \< 3 min | | **CSAT (AI-handled calls)** | Post-call satisfaction | > 3.5/5 | | **Escalation rate** | % transferred to human | \< 35% | Traditional IVR abandonment rates are typically 25–40%. AI IVR abandonment drops to 5–12% because callers aren't trapped in menus — they can say what they need and get an answer. ## Next Steps * [Cloud IVR guide](https://www.vobiz.ai/docs/solutions/cloud-ivr.md) — full IVR setup with VobizXML * [AI receptionist guide](https://www.vobiz.ai/docs/blogs/what-is-an-ai-receptionist.md) — full AI-first inbound handling * [Connect Retell AI](https://www.vobiz.ai/docs/integrations/retellai-dashboard.md) for no-code AI IVR setup * [XML IVR example](https://www.vobiz.ai/docs/examples/vobiz-ivr-xml-python.md) — working Python code * [Get a Vobiz number](https://console.vobiz.ai) to start This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.