> ## Documentation Index > Fetch the complete documentation index at: https://docs.getsmartalex.com/llms.txt > Use this file to discover all available pages before exploring further. # Replacing Your IVR > How SmartAlex replaces a traditional press-1-for-sales IVR or a live receptionist with a natural conversation. Most customers arrive with one of two setups: a traditional DTMF IVR ("press 1 for sales, 2 for support") or a live receptionist who answers and routes calls. SmartAlex replaces either, or both. ## The problem with traditional IVRs A DTMF IVR forces the caller to navigate your company's org chart using the numeric keypad. It's cheap, it's always been there, and it's universally disliked. | Metric | Industry average | | ----------------------------------------------------- | ---------------- | | Abandonment rate | 30-40% | | Misrouting rate | 15-25% | | Callers who press "0 for operator" to bypass the menu | \~45% | | Average handle time inflation vs. direct routing | 60-90 seconds | These numbers don't improve with a longer menu tree. They get worse. ## The problem with a live receptionist A human receptionist handles nuance beautifully. They also: * Cost between R20,000 and R40,000 per month in South Africa (R240k, R480k/year loaded) * Cover only business hours (or you pay for out-of-hours) * Take lunch, sick days, and leave * Miss calls during peak times * Can only take one call at a time * Need training to know every team member and their specialty * Struggle to remember every repeat caller A great receptionist is a genuine competitive advantage. Most businesses cannot afford a great one around the clock. ## What SmartAlex does differently Instead of a DTMF tree, the AI has a conversation. Instead of a human's memory, it has access to your CRM, knowledge base, calendar, and any tools you expose. It works 24/7, handles every call simultaneously, and remembers every caller. ```mermaid theme={null} flowchart TB subgraph Old["Traditional IVR"] I1[Caller dials] I2["Press 1 for Sales
Press 2 for Support
Press 3 for Billing"] I3[Caller presses button] I4[Queue or extension] I5[40% hang up before reaching anyone] I1 --> I2 --> I3 --> I4 I3 -.->|frustrated| I5 end subgraph New["SmartAlex"] S1[Caller dials] S2["AI: 'Hi, how can I help?'"] S3[Caller explains in natural language] S4["AI understands intent,
checks tools, either resolves
or transfers with context"] S5[Extension, queue, or resolution] S1 --> S2 --> S3 --> S4 --> S5 end style Old fill:#fee2e2 style New fill:#dcfce7 ``` ## A real example, medical practice A multi-practitioner medical clinic. 200 inbound calls per day. Previous IVR: ``` Welcome to Acme Medical. Press 1 for appointments Press 2 for billing Press 3 for Cardiology Press 4 for Pediatrics Press 0 for reception ``` **Before SmartAlex:** * 40% of callers hung up before finishing the menu * 25% pressed the wrong option and had to be re-routed * Reception handled \~200 calls/day during business hours only * No after-hours coverage * 3-minute average call handle time (menu + wait + human) **After SmartAlex (measured over 60 days):** * Abandonment: **8%** * Misrouting: **3%** * Reception handles **80 calls/day** (AI resolved the other 60%, transferred the remaining 40% with full context) * 24/7 coverage, 38 after-hours calls per week captured that would previously have gone to voicemail * 45-second average handle time * Patient satisfaction (post-call NPS) rose from 6.2 to 8.4 The AI handled standard appointment bookings, prescription refill requests, and general enquiries without ever transferring to a human. Clinical questions were transferred to the on-duty nurse with the reason for the call already summarised in the CRM note. ## A real example, BMW dealership (receptionist replacement) This is a different pattern. The customer doesn't have a DTMF IVR, they have a receptionist who knows every customer by name and routes calls to their assigned Sales Executive or Service Advisor. SmartAlex replicates this concierge model: ```mermaid theme={null} sequenceDiagram autonumber participant Customer participant AI as SmartAlex participant CRM participant SalesExec as James (Sales Exec) Customer->>AI: Calls main number AI->>CRM: Lookup caller by phone CRM-->>AI: "Craig Miller, owns a 2024 X5,
Sales Exec: James Thompson" AI->>Customer: "Good morning Craig, how can I help?" Customer->>AI: "Is my service appointment still on for Thursday?" AI->>CRM: Check service bookings CRM-->>AI: Thursday 10:30 confirmed AI->>Customer: "Yes, Thursday at 10:30. Anything else?" Customer->>AI: "Can I speak to James?" AI->>CRM: Check James's availability CRM-->>AI: In a meeting until 11am AI->>Customer: "He's in a meeting until 11. Want me to
take a message or get him to call you back?" Customer->>AI: "Ask him to call me" AI->>CRM: Create callback task for James AI->>Customer: "Done. He'll call you after 11." Note over AI: SMS confirmation sent ``` The AI recognised the caller, answered a substantive service question directly, and when the caller asked for a specific person, it didn't blindly transfer, it checked availability and managed the handoff intelligently. ## How the replacement is configured Three configuration decisions map your existing setup to SmartAlex: ### 1. What replaces the IVR menu? Instead of a menu, you give the AI a **first message** and a **system prompt** that describes who to help with what. The AI uses natural conversation to understand intent. Example first message for a medical practice: > "Good morning, Acme Medical. How can I help you today?" ### 2. What replaces the routing targets? Your **PBX extension directory**. You tell SmartAlex what extensions exist, who's on each, and what aliases callers might use ("sales team", "support", "John", "John Smith"). Example directory: | Extension | Display name | Owner | Aliases | | --------- | ------------ | ----------- | ------------------------------------- | | 101 | Reception | , | reception, front desk | | 102 | Sales | Sarah Jones | sarah, sales, sarah jones, sales team | | 103 | Support | John Smith | support, tech, john, help | | 104 | Billing | , | billing, accounts, invoices | When a caller says "put me through to Sarah in sales", the AI matches "Sarah" against the alias list, resolves to extension 102, and issues a SIP REFER. ### 3. What replaces the receptionist's judgement? Your **system prompt** plus **tools**. The prompt describes how to handle common requests. Tools give the AI access to calendar, CRM, knowledge base, SMS, and anything else you expose. Example system prompt fragment: > "For service appointments, always check the calendar first. For billing questions, transfer to Extension 104. For urgent clinical questions after hours, transfer to the on-call mobile number. Never offer medical advice, always transfer clinical questions to a qualified practitioner." ## What callers notice * The AI answers in one or two rings * It has your company's name and an appropriate greeting * It understands natural language, no keypad required * It handles common questions directly * Transfers are fast and land on the right person the first time * Recent callers are recognised by phone number when their contact exists in your CRM ## What callers don't notice (but should) * It's an AI. Callers that ask "am I talking to a robot?" are told the truth. Everyone else is just helped. * It never says "press 1", because DTMF trees are the problem we're replacing. * It doesn't read disclaimers, read back long menus, or apologise for wait times. It just helps. ## Next steps Six real industry examples with full scripts and configuration recipes. How to roll out, parallel-run, and cut over with zero downtime.