Glossary · 40 terms
Bots, flows & AI
Automation on WhatsApp: rule-based chatbots and AI agents, intents and retrieval, WhatsApp Flows and their data endpoints, guardrails, and the moment a conversation should move to a person.
Foundation· 10
Start here — the concepts everything else builds on.
- AI HallucinationAI hallucination is when a language model states something plausible but false — a wrong price, an invented refund policy — as confidently as a fact.Read definition
- Auto-ReplyAn instant, rule-triggered WhatsApp response — a greeting, an away or out-of-hours message, or a keyword reply — sent without a human or a full chatbot.Read definition
- ChatbotSoftware that converses with customers on WhatsApp — through buttons and menus, keyword rules or AI — answering questions and completing simple tasks.Read definition
- Intent RecognitionIntent recognition classifies what a customer wants from a message, such as tracking an order or a refund, with a confidence score that picks the handler.Read definition
- Knowledge BaseThe curated FAQs, policies, help articles and product facts that bots, AI agents and human agents answer customers from — owned, dated and kept current.Read definition
- Large Language ModelLLMA transformer model trained on vast text that writes a reply one token at a time — the engine beneath AI chatbots and agents, not an agent itself.Read definition
- Natural Language UnderstandingNLUThe layer that turns a customer's free text into structured meaning — language, intent, entities and sentiment — that a bot or routing rule can act on.Read definition
- No-Code Chatbot BuilderA visual drag-and-drop canvas for building WhatsApp bot journeys from triggers, messages, conditions and API blocks, without writing code.Read definition
- WhatsApp FlowsMulti-screen native forms that open inside a WhatsApp chat, letting customers book, sign up or answer surveys without leaving the conversation.Read definition
- Workflow AutomationEvent-driven rules that connect WhatsApp to your store, CRM and back office: when something happens, check conditions, then send or update automatically.Read definition
Practitioner· 22
What you need once you are running it day to day.
- Agent AssistAI that works beside a human agent in a live chat, suggesting replies, summarising threads and fetching facts while the agent decides what to send.Read definition
- AI AgentAn LLM-driven system that plans and takes actions through tools over several steps — checking an order, booking a slot — instead of only answering.Read definition
- AI GuardrailsChecks placed before and after an LLM — topic limits, PII filters, policy and tone rules, action caps — that keep a bot's replies on-brand and in policy.Read definition
- Containment RateThe share of conversations a bot resolves without a human taking over — measured honestly, it excludes abandoned chats and customers who come back.Read definition
- Context WindowThe maximum number of tokens a language model can take into account at once — system prompt, chat history, retrieved text and its own reply combined.Read definition
- Conversation State ManagementStoring where each customer is in a bot journey — current step, answers so far, last intent — so stateless webhook calls can pick up the thread.Read definition
- Conversational ComponentsPer-number WhatsApp settings — ice breakers, slash commands and the welcome-message event — that give customers ready-made ways to start a chat.Read definition
- Entity ExtractionEntity extraction pulls structured values such as order IDs, dates, PIN codes and amounts out of a customer's free-text message so a bot can act on them.Read definition
- Fallback IntentA fallback intent is what a chatbot does when no intent clears its confidence threshold: ask to rephrase, offer a menu, or escalate after repeated misses.Read definition
- Flow LifecycleThe Flow lifecycle is the set of statuses a WhatsApp Flow moves through — Draft, Published, Throttled, Blocked, Deprecated — and what you can do in each.Read definition
- Flow Response MessageA Flow response message is the inbound nfm_reply your webhook receives when a customer completes a WhatsApp Flow, carrying response_json and the flow_token.Read definition
- Human HandoffHuman handoff passes a live WhatsApp conversation from a chatbot to a human agent, with the transcript and details, so the customer never repeats themselves.Read definition
- Language DetectionIdentifying the language and script of a message — Hindi, Tamil, or Hindi typed in Latin letters — so a bot replies in kind and picks matching content.Read definition
- Lead QualificationUsing bot questions or WhatsApp Flows to score inbound leads on need, budget, timeline and location, then routing them before a salesperson engages.Read definition
- Meta Business AgentMeta's own AI agent that answers a business's customers on WhatsApp from its information, FAQs, website and catalogue, billed per token since August 2026.Read definition
- Prompt InjectionAn attack in which text a bot reads — a customer message, a document or a tool result — carries instructions that override the bot's own rules.Read definition
- Retrieval-Augmented GenerationRAGA technique that retrieves passages from your own documents at question time and adds them to an LLM's prompt, so answers are grounded and citable.Read definition
- Sentiment AnalysisScoring each customer message as positive, neutral or negative — and spotting frustration — so support teams can prioritise, escalate or follow up.Read definition
- Speech-to-TextSTTSoftware that converts spoken audio — WhatsApp voice notes and calls — into text that bots, search and support agents can read and act on.Read definition
- Structured Output (JSON Mode)Constraining an LLM to reply in JSON that matches a schema you define, so your code can turn it into buttons, template variables or CRM fields.Read definition
- System PromptThe standing instructions an LLM receives before any customer message, setting a bot's role, scope, tone, rules, tools and reply format.Read definition
- Tool Calling (Function Calling)An LLM feature in which the model replies with a structured request to run one of your functions — such as check_order_status — instead of text.Read definition
Expert· 8
Edge cases, internals and the details that matter at scale.
- Fine-TuningFine-tuning further trains a pretrained language model on your own examples so it adopts a tone, format or task behaviour. It changes the model's weights.Read definition
- Flow EndpointA Flow endpoint is the HTTPS server a WhatsApp Flow calls to load or validate data mid-flow, with every request and response encrypted using RSA and AES-GCM.Read definition
- Flow JSONFlow JSON is the versioned document that defines a WhatsApp Flow: its screens, components, routing between screens and actions such as navigate and complete.Read definition
- Interactive Flow TokenThe flow_token string a business sets when it sends a WhatsApp Flow and gets back on every endpoint call and on completion, used to tie the session to a record.Read definition
- LLM EvaluationOffline testing of a bot or AI agent against a fixed set of real cases, scored by code checks, rubrics and LLM judges, before a prompt or model change ships.Read definition
- Model Context ProtocolMCPAn open JSON-RPC protocol that lets LLM applications discover and call tools, read resources and use prompts exposed by MCP servers in one standard way.Read definition
- Vector DatabaseA database that stores embeddings and quickly finds the most similar ones, using approximate nearest-neighbour indexes such as HNSW or IVF.Read definition
- Vector EmbeddingsLists of numbers that encode the meaning of text so similar meanings sit close together, letting software match questions to answers by meaning, not words.Read definition
Put these terms to work.
Go live on the official WhatsApp Business API in 2 hours with a 7-day free trial — no card required.
7-day free trial Enterprise-grade reliability Live in 2 hours Built for Indian businesses