Glossary: AI assistant terms

40 terms

• Data retention. Data retention is the period for which the provider stores the conversations and the data they contain.
• AI assistant. An AI assistant is software that writes answers to the questions of customers or colleagues from the company’s own content.
• AI disclosure (AI Act, Article 50). AI disclosure is the mandatory notice that the customer is talking to artificial intelligence and not to a human.
• AI answer engine visibility. AI answer engine visibility means that search tools such as ChatGPT cite the company’s page when a customer asks them a question.
• Handoff to a human. The handoff is the step in which the assistant asks for the customer’s contact details and forwards the question to a colleague.
• Onboarding. Onboarding is the process in which the knowledge base is filled, the tone and the handoff are set up, and the widget goes onto the website.
• Chatbot. A chatbot is a program that answers on a website in a conversational way; the AI-based version writes from the company’s content, the rule-based one gives pre-written…
• Multi-assistant discount. The multi-assistant discount is a percentage discount given for subscribing to several assistants together.
• Go-live. Go-live is the moment when the assistant answers real visitors on the real website.
• Lead form. The lead form is the short field in the chat window where the customer enters their contact details when their question needs a human.
• EU data storage. EU data storage means that the conversations, the knowledge base and the model calls all run on servers within the European Union.
• Source citation. A source citation is the indication shown next to the answer of which document or page the assistant took the information from.
• GDPR and data processor. The provider of the AI assistant is the data processor of customer data, and a data processing agreement (DPA) must be concluded with it.
• Installation via Google Tag Manager. Installation via Tag Manager means that the widget’s code snippet is inserted into Google Tag Manager instead of the website template.
• FAQ bot. An FAQ bot picks from pre-written question-and-answer pairs based on keywords; it does not write new sentences.
• Hallucination. Hallucination is when an AI confidently gives an answer that is not in its sources.
• Tone of voice setting. The tone of voice setting determines the assistant’s style (short or detailed, matter-of-fact or friendly).
• Weekly summary. The weekly summary is the email that arrives on Monday morning and shows the previous week’s questions and results without logging in.
• Allowance and overage. The allowance is the monthly number of units included in the package; overage is what happens once the allowance is used up.
• Context and document match. A match is the part of the knowledge base that the assistant retrieves for the question and passes to the model to answer.
• llms.txt. llms.txt is a simple text file in the root of a website that summarises for AI systems who the company is and what each page is about.
• Answered question. The answered question is the unit in which use of the assistant is measured: one customer question that received a substantive answer.
• Question outside business hours. An after-hours question is a customer question that arrives between 17:00 and 8:00 on weekdays or at the weekend.
• Net and gross price. The net price is the price without VAT, the gross price is the price including VAT; business services are often priced net.
• Language model. The language model is the artificial intelligence that understands the question and formulates the answer; the company’s knowledge comes from the knowledge base.
• Test question. A test question is a customer question prepared in advance to check before go-live that the assistant answers correctly.
• Order status lookup. An order status lookup is when the assistant tells the customer where their parcel is, based on the order number.
• Shoprenter integration. The Shoprenter integration connects the product data and the widget of a Shoprenter online store to the assistant.
• Structured data (schema). Structured data is a machine-readable description embedded in the website of what the page contains (company, product, price, FAQ).
• Rule-based and generative assistant. A rule-based assistant follows a decision tree; a generative one writes a new answer from the content.
• Formal and informal address setting. The formal and informal address setting decides whether the assistant addresses the customer informally or formally.
• Product sync. Product sync is the regular transfer of the online store’s product data (name, price, stock, description) into the knowledge base.
• Forbidden topics. Forbidden topics are the subject areas in which the assistant does not answer; it hands the question over to a human.
• Multilingual answers. Multilingual answers means that the assistant answers in the customer’s language, even if the knowledge base is in another language.
• Knowledge base. The knowledge base is the collection of the company’s documents and pages from which the AI assistant answers.
• Knowledge base upload. Knowledge base upload is the reading in of the website, the documents and the product data for the assistant.
• UNAS integration. The UNAS integration connects the product data and the widget of an UNAS online store to the assistant.
• Response time. Response time is the time from sending the question to the answer appearing, including the search and the model run.
• Widget and code snippet. The widget is the chat window shown on the website; the code snippet is the few lines that must be inserted into the website for the widget to appear.
• WooCommerce plugin. The WooCommerce plugin connects the product data and the widget of a WordPress-based online store to the assistant.

Frequently asked questions

Glossary: AI assistant and chatbot terms