Plain-English glossary
Every term you'll hit while shopping for or building an AI chatbot, explained without the vendor fluff.
Agentic AI, What It Means for Real Chatbot Builds
Agentic AI describes systems that plan and take multi-step actions toward a goal. Here is what that buys you in a chatbot and what it costs you.
Barge-In
Letting a caller interrupt your AI agent mid-sentence, why it is non-negotiable on phone calls, and the tuning that keeps it from misfiring.
Context Window in AI Chatbots, Explained
The context window is the total amount of text a model can hold in mind at once. Here is why it decides what your chatbot remembers and what it forgets.
Conversation Design for Chatbots and Voice Agents
Conversation design is the craft of deciding what your bot says, asks and does. It is the work that separates a useful agent from an annoying one.
DID Number (Direct Inward Dialing)
The phone number your AI agent answers on, how to choose between a new number and your existing one, and the porting trap to avoid.
DTMF, the Keypad Tones Your AI Agent Still Needs
DTMF is the touch-tone signal sent when a caller presses a key. Here is why AI voice agents still need it for menus, extensions and verification.
Dynamic Variables in AI Chatbots and Voice Agents
Dynamic variables inject live data like caller ID or business hours into a prompt at runtime. Get them wrong and your agent goes silent on answer.
Fine-Tuning, and Why Most Chatbots Do Not Need It
Fine-tuning retrains a model on your own examples. Here is what it actually fixes, what it does not, and why prompting plus retrieval usually wins.
Function Calling, How a Chatbot Actually Does Things
Function calling lets a chatbot run real actions like booking, lookups and transfers. Here is how it works and where it breaks in production.
Guardrails
The rules and hard blocks that stop an AI agent doing something it should not, and why the ones that matter live outside the prompt.
Hallucination
When a chatbot states something false with full confidence, why it happens, and the prompt pattern that killed the worst case I have shipped.
Human Handoff, Getting the Customer to a Person
Human handoff is when a chatbot passes the conversation to a real person. Designing it well matters more than almost anything else in the build.
Intent
What a user is actually trying to do, how intent classification used to drive chatbots, and how much of it LLMs made unnecessary.
IVR (Interactive Voice Response)
The press-one-for-sales phone menu, how it compares to an AI voice agent, and why the old IVR is usually still sitting in front of your new one.
Knowledge Base
The curated set of documents a chatbot can look things up in, how it differs from the prompt, and how to keep it from going stale on you.
Latency
The delay between a caller finishing a sentence and the agent replying, why it decides whether a voice agent feels real, and where to find the milliseconds.
LLM (Large Language Model)
What a large language model actually is, how it powers a chatbot or voice agent, and what changes for you when you pick one over another.
MCP (Model Context Protocol) for Chatbot Builders
MCP is a standard way to expose tools and data to AI models. Here is what it changes for chatbot builds and where it is still early as of mid-2026.
Prompt Injection, the Security Risk in Every Chatbot
Prompt injection is when untrusted text hijacks your chatbot's instructions. Here is how it happens in real deployments and how to limit the damage.
PSTN, the Real Phone Network Behind Your AI Agent
The PSTN is the traditional public telephone network. Here is why it still matters when you put an AI voice agent on a real business phone number.
RAG (Retrieval-Augmented Generation)
How a chatbot looks up your real content before answering, why it beats stuffing facts into the prompt, and where retrieval quietly fails.
Session Memory, What Your Chatbot Remembers and When
Session memory is what a chatbot retains during and between conversations. Getting the boundary right decides whether it feels sharp or forgetful.
SIP Trunking
The internet plumbing that carries business phone calls, why it matters for AI voice agents, and the part of it you actually have to care about.
STT (Speech-to-Text)
The transcription layer of a voice agent, why most misunderstood calls start here, and what to check before blaming the AI model.
System Prompt
The standing instructions that define a chatbot's job, tone, and limits, and why it is the single highest-leverage file in the whole build.
Token, the Unit AI Chatbots Bill and Think In
A token is a chunk of text roughly three quarters of a word. Tokens set what your chatbot costs, how fast it replies, and how much it can hold.
TTS (Text-to-Speech)
The layer that turns your agent's words into a voice, what actually drives quality on a phone call, and how voice choice affects cost and latency.
Turn-Taking
How a voice agent decides when the caller has finished speaking and when it should talk, and why this is the hardest part of sounding natural.
Voice Cloning for AI Phone Agents, Explained
Voice cloning builds a synthetic voice from recordings of a real person. Here is when it helps an AI receptionist and when a stock voice is the better call.
Webhook, the Glue Behind Most Chatbot Integrations
A webhook is an automatic HTTP message sent when something happens. It is how chatbots trigger CRM updates, SMS follow-ups and missed-call flows.