Skip to main content

Core Concepts

Before diving into the platform, it's essential to understand the fundamental concepts that power Hanc.AI. This section explains everything from basic AI principles to specific platform terminology.

What is Artificial Intelligence (AI)?

Artificial Intelligence is computer software that can perform tasks that typically require human intelligence. Unlike traditional software that follows exact rules, AI can:

  • Understand natural language — comprehend human speech and text
  • Learn patterns — improve based on data and examples
  • Make decisions — choose appropriate responses based on context
  • Generate content — create human-like text, speech, or other outputs

AI in Simple Terms

Think of AI as a very sophisticated pattern-matching system:

Traditional Software:
IF customer says "hello" THEN respond "Hello, how can I help?"
IF customer says "hi" THEN respond "Hello, how can I help?"
IF customer says "hey" THEN respond "Hello, how can I help?"
... (needs a rule for every possible input)

AI-Powered Software:
UNDERSTAND that "hello", "hi", "hey", "good morning", "howdy"
all mean a greeting → respond appropriately based on context

AI doesn't need explicit rules for every situation — it understands intent and context.


What is a Voice Agent?

A Voice Agent (also called a voice assistant or voice bot) is an AI-powered program that communicates with people through spoken conversation.

How Voice Agents Work

┌─────────────────────────────────────────────────────────────────┐
│ VOICE AGENT PIPELINE │
└─────────────────────────────────────────────────────────────────┘

Customer speaks AI processes Agent responds
│ │ │
▼ ▼ ▼
┌───────────┐ ┌───────────┐ ┌───────────┐ ┌───────────┐
│ Voice │───▶│ Speech │───▶│ AI │───▶│ Text to │
│ Input │ │ to Text │ │ Brain │ │ Speech │
│ │ │ (STT) │ │ (LLM) │ │ (TTS) │
└───────────┘ └───────────┘ └───────────┘ └───────────┘


┌───────────────────┐
│ Knowledge Base │
│ (Your business │
│ information) │
└───────────────────┘

Step by step:

  1. Customer speaks — "What time do you close on Saturdays?"
  2. Speech-to-Text (STT) — Converts audio to text
  3. AI Brain (LLM) — Understands the question, searches Knowledge Base
  4. Knowledge Base — Contains your business hours: "Saturday: 10am-6pm"
  5. AI generates response — "We're open on Saturdays from 10am to 6pm"
  6. Text-to-Speech (TTS) — Converts text back to natural-sounding speech
  7. Customer hears — Natural voice response

Voice Agent vs. Chatbot

FeatureVoice AgentChatbot
CommunicationSpoken voiceWritten text
ChannelPhone calls, voice widgetsWebsite chat, messaging apps
InteractionReal-time conversationCan be asynchronous
ComplexityHigher (speech processing)Lower
Use casesPhone support, appointmentsWebsite support, FAQ

Hanc.AI specializes in Voice Agents — AI that handles real phone conversations.


Large Language Models (LLM)

The "brain" of a voice agent is a Large Language Model (LLM). This is the AI technology that:

  • Understands what customers are saying
  • Formulates appropriate responses
  • Maintains conversation context
  • Follows your instructions (the "prompt")
ProviderModelsCharacteristics
OpenAIGPT-4, GPT-4o, GPT-4o-miniIndustry standard, excellent general performance
AnthropicClaude 3.5 Sonnet, Claude 3 OpusStrong reasoning, safety-focused
GoogleGemini Pro, Gemini FlashFast, good multilingual support
MetaLlama 3Open-source alternative

Hanc.AI supports multiple LLM providers, allowing you to choose the best fit for your needs.

How LLMs "Think"

LLMs don't truly "think" — they predict the most likely next words based on:

  1. Training data — Billions of text examples from the internet
  2. Your prompt — Instructions you give about how to behave
  3. Knowledge Base — Your specific business information
  4. Conversation history — What was said earlier in the call
                 ┌────────────────────────────┐
│ LLM Decision │
└────────────────────────────┘

┌─────────────────────┼─────────────────────┐
│ │ │
▼ ▼ ▼
┌───────────────┐ ┌───────────────┐ ┌───────────────┐
│ Your Prompt │ │ Knowledge │ │ Conversation │
│ (behavior) │ │ Base │ │ Context │
│ │ │ (facts) │ │ (history) │
└───────────────┘ └───────────────┘ └───────────────┘

Knowledge Base (RAG)

A Knowledge Base is a collection of documents containing information about your business. The AI uses this information to answer customer questions accurately.

What is RAG?

RAG stands for Retrieval-Augmented Generation:

  • Retrieval — Finding relevant information from your documents
  • Augmented — Adding that information to the AI's context
  • Generation — Creating a response using that information

Why RAG Matters

Without RAG, the AI only knows general information from its training. It doesn't know:

  • Your specific services and prices
  • Your business hours
  • Your policies
  • Your team members
  • Anything unique to your business

With RAG:

Customer: "How much does a haircut cost?"
AI: [Searches Knowledge Base] → Found: "Men's haircut: €25, Women's haircut: €35"
AI: "A men's haircut is €25, and a women's haircut is €35."

Without RAG:

Customer: "How much does a haircut cost?"
AI: "I don't have information about your specific prices.
Generally, haircuts range from €15 to €50..."

Knowledge Base Content

Your Knowledge Base should include:

CategoryExamples
Services/ProductsDescriptions, prices, duration
Business InfoHours, location, contact details
PoliciesCancellation, refunds, booking rules
FAQCommon questions and answers
StaffTeam members, specializations
PromotionsCurrent offers, discounts

The Prompt

A Prompt is a set of instructions that tells the AI how to behave. It's like giving an employee their job description and guidelines.

Prompt Components

┌─────────────────────────────────────────────────────────────────┐
│ AGENT PROMPT │
├─────────────────────────────────────────────────────────────────┤
│ │
│ ROLE │
│ "You are Sophie, a virtual assistant for ABC Dental Clinic" │
│ │
│ GOALS │
│ - Answer questions about services │
│ - Book appointments │
│ - Provide clinic information │
│ │
│ RULES │
│ - Only use information from Knowledge Base │
│ - Never make up information │
│ - Be polite and professional │
│ - Transfer to human if customer is upset │
│ │
│ PERSONALITY │
│ - Friendly but professional │
│ - Speak in short, clear sentences │
│ - Use the customer's name when known │
│ │
└─────────────────────────────────────────────────────────────────┘

Prompt Best Practices

DoDon't
Be specific and clearUse vague instructions
Define boundariesLeave behavior undefined
Include examplesAssume AI understands context
Set escalation rulesLet AI handle everything
Specify language and toneForget communication style

Speech Technologies

Speech-to-Text (STT)

Converts spoken words into written text. Also called:

  • Automatic Speech Recognition (ASR)
  • Voice Recognition
  • Transcription

Quality factors:

  • Accent handling
  • Background noise resistance
  • Technical vocabulary recognition
  • Real-time vs. batch processing

Text-to-Speech (TTS)

Converts written text into spoken words. Modern TTS sounds natural, with:

  • Multiple voice options (male, female, different ages)
  • Emotion and intonation
  • Multiple languages
  • Adjustable speed

Voice selection considerations:

  • Match your brand (professional, friendly, energetic)
  • Language and accent requirements
  • Customer demographics

Real-Time Communication (WebRTC)

WebRTC (Web Real-Time Communication) is the technology that enables:

  • Browser-based voice calls
  • Low-latency audio streaming
  • No plugins required

Hanc.AI uses LiveKit — a WebRTC platform — for:

  • Voice widget on your website
  • Test calls in the dashboard
  • Real-time audio processing

Key Platform Terms

TermDefinition
AgentA configured AI voice assistant
WorkspaceA container for agents, belonging to one organization
CustomerAn organization using the platform (your business)
Call LogRecord of a conversation between agent and caller
ToolAn action the agent can perform (book appointment, transfer call)
ActionA specific capability enabled for an agent
First MessageThe greeting the agent says when answering
FallbackWhat happens when the agent can't help (transfer to human)
EscalationTransferring a call to a human operator

Next Steps

Now that you understand the core concepts:

  1. How It Works — Technical architecture overview
  2. Account Setup — Create your account
  3. Your First Agent — Build your first voice agent

Summary

ConceptKey Point
AISoftware that understands and generates human-like responses
Voice AgentAI that communicates through spoken conversation
LLMThe "brain" that processes language and generates responses
Knowledge BaseYour business information that the AI uses to answer questions
RAGTechnology that retrieves relevant info from your documents
PromptInstructions that define the agent's behavior
STT/TTSTechnologies for converting between speech and text

Understanding these concepts will help you configure more effective voice agents and troubleshoot issues when they arise.