Glossary

Short explanations of the terms used in the HaveAI panel and guides.

Assistant and Training

Assistant / AI agent: The AI assistant that answers customer questions with the business information and behavior instructions defined for it.

Training: The process of adding and processing your business information as sources the assistant can use. This use in HaveAI does not mean training the base AI model from scratch.

Knowledge base: The whole of the sources the assistant makes use of, such as FAQs, free-form information, products, services, documents and web content.

FAQ: An answer prepared for a specific frequently asked question. For example, the answer to "How many days is the return period?"

Free-form information: Content that explains a topic without being tied to a question-answer format. For example, delivery areas and delivery conditions explained together.

RAG: The assistant searching knowledge sources while preparing an answer and making use of relevant content. It is the abbreviation of the English phrase "Retrieval-Augmented Generation."

Embedding: A numerical representation used to compare the semantic similarity of texts.

Semantic search: A search method that tries to find information with similar meaning even when the same words are not used.

Knowledge gap: A question or topic where the assistant is thought to not find enough information. It is reviewed to improve training content.

Training processing allowance: The usage allowance of the text processed during training, measured by character. It differs from the storage amount of saved documents.

Conversations and Channels

Customer: A person who contacts your business through supported channels.

Conversation / chat: The correspondence between the customer and the assistant or team members.

Account session: The sign-in state that lets you access your account after signing in to HaveAI. It is not the same as a customer conversation.

Channel: The communication medium messages come from. The web widget, Telegram, WhatsApp, Instagram DM, Facebook Messenger and Slack are examples.

Widget: The chat component added to your website. The visitor sends messages through this component; the answer can be given by the assistant or the team member who takes over the conversation.

Zernio: The service that mediates HaveAI's connection to Meta channels and the delivery of messages.

Human handoff: The process of transferring a conversation from the AI assistant to a team member. The handoff request being created and the agent taking over the conversation are separate stages.

Greeting questions / intake: The name, email or need-related information requested from the visitor before the chat in the widget.

Sentiment analysis: The AI assessment of a message's positive, neutral or negative tone. It is not an exact measurement of the person's real emotion.

Appointment request: The customer's wish to have a meeting or receive a service at a specific time.

Appointment record: The appointment created in the system with the service, date, time and related information. It differs from the request being expressed.

Automation: A rule that starts an action when a specific event or condition occurs.

Template: Saved text team members can use to reply quickly.

Support ticket: The correspondence record maintained with the HaveAI support team about a specific issue or question.

Connections and Security

API: The interface that lets your own software communicate with HaveAI through specific operations.

API key: The secret credential used for API access. HaveAI keys have the hv_live_sk_ or hv_test_sk_ prefix.

Widget token: The setup identity that matches the widget on the website with the relevant account or agent. It appears in the setup code; it is not the same as the secret API key.

Scope: The permission that determines which operations an API key can access.

Webhook: The connection method that sends an automatic notification to another system when a specific event occurs.

PII: Information that can be associated with a person. A name, email, phone or ID number are examples. HaveAI can detect and mask certain data formats.

2FA: Two-factor authentication. Using an additional verification step at sign-in.

TOTP: A one-time code that changes over time. The six-digit codes authenticator apps produce use this method.

DPA: Data Processing Agreement. It regulates the terms for HaveAI processing personal data on behalf of your business.

Related Topics