100 terms · A to Z
Glossary
Clear definitions of the terms used in AI and customer communication, from agents and knowledge to handoff and telephony.
Look up the term you ran into.
A
A/B testing
A/B testing means showing two different variants, for example two different greeting phrases for an AI agent, to different visitors and measuring which performs best.
See also: conversion rate optimization (CRO), test environment
MarketingLinkAI agent
An AI agent is a digital AI colleague that understands the customer's question and context, answers or carries out the right action, and brings in a person when needed. At Aveexia, one AI agent handles customer communication across web chat, email, SMS, WhatsApp and voice.
See also: AI phone agent, chatbot
Product / categoryLinkAI onboarding
AI onboarding is the process of setting up an AI agent for a new business: filling the knowledge base, connecting integrations and adjusting the tone of voice. The simpler the onboarding, the faster the business can start using the AI agent for real.
See also: knowledge base, sandbox environment
ProductLinkAI phone agent
An AI phone agent answers incoming calls with an AI voice, understands what the caller wants and can book appointments, answer questions or transfer the call to a staff member. Unlike a traditional switchboard, it understands natural speech instead of keypad menus.
See also: AI agent, phone switchboard
Product / categoryLinkAI summary
AI summarisation is technology that automatically condenses a longer conversation or text into a short summary, for example so a staff member can quickly see what a call was about before taking it over.
See also: transcription, live handoff
TechnologyLinkAPI integration
An API integration lets two systems exchange information and carry out agreed actions. Aveexia connects your systems during onboarding.
See also: webhook, CRM integration
TechnologyLinkAPI key
An API key is a secret credential a client uses to identify itself to an API. Treat it like a password: grant only the permissions it needs, keep it on a server, and do not expose it in browser code, URLs or analytics. Revoke or rotate the key if it is exposed.
See also: API integration, role-based access
TechnologyLinkaudit log
An audit log records who did what and when in a system, making it possible to trace changes and detect unauthorised activity after the fact.
See also: role-based access, encryption
SecurityLinkautomatic response
An automatic response is a predefined or AI-generated reply sent without a human writing it manually each time. The foundation for how an AI agent answers recurring questions quickly.
See also: SMS automation, email automation
ProductLinkaverage handling time (AHT)
AHT (average handling time) is the average time it takes to resolve a customer case, from first contact to closure. An AI agent can lower AHT by resolving simple cases directly without a staff member getting involved.
See also: first response time (FRT), customer satisfaction (CSAT)
MeasurementLink
B
booking system
A booking system is the software a business uses to manage appointments and meetings. Aveexia connects to your existing booking system through an integration, so the AI agent can book, reschedule and cancel directly.
See also: calendar integration, API integration
ProductLinkbounce rate
Bounce rate is the share of visitors who leave a website after viewing only a single page, without further interaction. An AI agent that quickly answers a visitor's question can lower bounce rate by keeping the visitor engaged.
See also: landing page, conversion rate
MarketingLink
C
calendar integration
A calendar integration connects Aveexia to your calendar (for example Google Calendar or Outlook) so booked meetings appear directly without manual entry, and available time slots are fetched in real time.
See also: booking system, API integration
ProductLinkcall flow
A call flow is the pre-programmed path a phone call follows through a system, which menu choices, queues or AI responses the caller encounters in sequence. Designed to route the caller to the right answer or person as quickly as possible.
See also: IVR (interactive voice response), phone switchboard
TelephonyLinkcall to action (CTA)
A CTA (call-to-action) is the button or text that prompts the visitor to take the next step ("Book a demo", "Try for free") and is decisive for how many actually convert.
See also: landing page, conversion rate
MarketingLinkchat widget
A chat widget is the small chat box, usually in the bottom-right corner, that visitors click to talk to an AI agent or chatbot on a website. Usually requires just a snippet of code pasted into the site.
See also: chatbot, web-based chat
ProductLinkchatbot
A chatbot is a program that carries on a text-based conversation with a user, typically through a chat widget on a website. Modern chatbots are built on conversational AI and can understand freely worded questions, not just pre-programmed button choices.
See also: conversational AI, chat widget
Product / categoryLinkchatbot flow
A chatbot flow is the logical structure behind a conversation, which questions are asked, in what order, and which paths the conversation can take depending on the answers. Modern AI chatbots build flows dynamically rather than following a fixed script.
See also: dialogue tree, conversation design
TechnologyLinkchatbot persona
A chatbot persona is the personality and tone of voice given to an AI agent (formal or informal, brief or detailed) so it feels like a natural part of the company's brand rather than a generic AI tool.
See also: prompt engineering, conversation design
ProductLinkchurn
Churn is the share of customers who stop using a service during a given period, for example by cancelling their subscription. One of the most important metrics for a SaaS company's long-term health.
See also: customer lifetime value (CLV), subscription model
BusinessLinkcloud service (SaaS)
SaaS (Software as a Service) is a cloud service used through the browser for a recurring fee, instead of installing and maintaining software yourself. Aveexia is delivered as a SaaS product.
See also: SaaS agreement, subscription model
TechnologyLinkcontext window
The context window is how much of an ongoing conversation (and knowledge base) a language model can "remember" and take into account at once when formulating a response. A larger context window lets the AI keep track of longer, more complex conversations.
See also: multi-turn conversation, LLM (large language model)
TechnologyLinkconversation design
Conversation design is the work of shaping how an AI conducts its dialogues (word choice, tone, response length and how it handles misunderstandings) so the conversation feels natural and helpful.
See also: chatbot persona, chatbot flow
TechnologyLinkconversational AI
Conversational AI is the umbrella term for technology that lets computers hold natural, dialogue-like conversations with people, in text or speech. It relies on language models (LLMs) and NLP to interpret intent and generate relevant responses.
See also: NLP (natural language processing), LLM (large language model)
TechnologyLinkconversion rate
Conversion rate is the share of visitors who complete a desired action (booking a meeting, filling in a form, trying the product) out of total visitors. One of the key metrics for judging whether an AI agent actually drives business value.
See also: call to action (CTA), conversion rate optimization (CRO)
MarketingLinkconversion rate optimization (CRO)
CRO (conversion rate optimization) is the systematic work of improving the share of visitors who complete a desired action, for example by testing different CTA copy or an AI agent's opening greeting.
See also: A/B testing, call to action (CTA)
MarketingLinkCRM integration
A CRM integration connects customer conversations to a customer database. You decide which fields are passed on.
See also: API integration, lead capture
TechnologyLinkcustomer lifetime value (CLV)
CLV (customer lifetime value) is the total revenue a customer is expected to generate over the entire relationship with a business. The lower the churn and the longer customers stay, the higher the CLV.
See also: churn, subscription model
BusinessLinkcustomer satisfaction (CSAT)
CSAT (customer satisfaction) measures how satisfied customers are with a specific interaction, usually through a simple question after a conversation ends. A common metric for evaluating an AI agent's quality.
See also: Net Promoter Score (NPS), first response time (FRT)
MeasurementLink
D
data processing agreement (DPA)
A data processing agreement (DPA) is a legal contract governing how a vendor like Aveexia may handle personal data on a company's behalf. Required under GDPR whenever a third party processes customer data.
See also: GDPR, data storage in the EU
LegalLinkdata residency
Data residency describes the country or region where data is stored or processed. It affects risk assessment, contractual requirements and transfer mechanisms, but EU storage alone is not a complete GDPR assessment.
See also: data storage in the EU, GDPR
LegalLinkdata storage in the EU
Data storage in the EU describes customer data being stored on servers within the EU/EEA. Storage region can affect risk assessment, contractual requirements and transfer mechanisms, but EU storage alone is not a complete GDPR assessment.
See also: GDPR, data residency
LegalLinkdialogue tree
A dialogue tree is a branching structure of predefined questions and answers, common in older, rule-based chatbots. Modern AI agents like Aveexia instead rely on language models that can respond flexibly outside a fixed tree.
See also: chatbot flow, conversation design
TechnologyLinkdigital worker
A digital employee is an AI that performs tasks a staff member would otherwise do, like answering the phone or chat, and is often described as part of the team rather than just a tool.
See also: virtual assistant, AI agent
ProductLink
E
email automation
Email automation is automatic emails sent based on an event or rule, for example a confirmation when an AI agent captures a new lead.
See also: automatic response, lead capture
ProductLinkencryption
Encryption converts data into an unreadable form that can only be unlocked with the right key, protecting information both in transit and at rest. A basic security measure for all customer data an AI agent handles.
See also: two-factor authentication, audit log
SecurityLinkentity recognition
Entity recognition is when an AI identifies and extracts specific pieces of information from a sentence (such as a date, a name or a service) in order to act on them, for example to book the correct time slot.
See also: intent recognition, NLP (natural language processing)
TechnologyLinkescalation
Escalation is when a case is lifted from the AI to a human, or to a more senior staff member, because it's too complex, sensitive or important for the AI to handle on its own.
See also: live handoff, sentiment analysis
ProductLink
F
fallback response
A fallback response is what an AI gives when it can't find a good enough answer in the knowledge base, for example offering to connect the visitor to a human instead of guessing. Good fallback handling is essential to avoid incorrect answers.
See also: live handoff, knowledge base
ProductLinkFAQ module
An FAQ module is a structured collection of common questions and answers that an AI agent uses as part of its knowledge base, and which is often also displayed as its own section on the website for SEO purposes.
See also: knowledge base
ProductLinkfine-tuning
Fine-tuning is further training an existing AI model on specific data so it becomes better at a narrow task or domain. Many AI agents rely on RAG (a knowledge base) instead of fine-tuning, since it's faster to update.
See also: training data, RAG (retrieval-augmented generation)
TechnologyLinkfirst response time (FRT)
FRT (first response time) is the time it takes from a customer reaching out until they get their first reply. A correctly configured AI agent can reduce FRT substantially outside office hours, but actual response time depends on the channel and runtime.
See also: average handling time (AHT), customer satisfaction (CSAT)
MeasurementLinkfreemium
Freemium is a business model where a basic version of a service is free, while more advanced features require a paid plan. A way to lower the barrier to trying a product.
See also: subscription model, SaaS agreement
BusinessLinkfunnel model
The funnel model describes the customer journey in stages, from someone first hearing about a business (awareness) to becoming a paying customer (conversion), often illustrated as a funnel since the number of people shrinks at each stage.
See also: lead nurturing, marketing qualified lead (MQL)
MarketingLink
G
GDPR
GDPR (the General Data Protection Regulation) governs how personal data may be collected, stored and used. An AI agent handling customer conversations needs a legal basis, data minimisation, transparency and appropriate agreements; storage region and transfers must be assessed separately.
See also: data processing agreement (DPA), data storage in the EU
LegalLink
H
hyperautomation
Hyperautomation is combining multiple automation technologies (AI, RPA, integrations) to automate entire business processes end to end, not just individual tasks.
See also: RPA (robotic process automation), API integration
MarketingLink
I
intent recognition
Intent recognition is the process where an AI interprets what a user is actually trying to achieve with their message (for example booking an appointment, asking about price, or cancelling) even when phrased differently each time. Essential for an AI agent to respond correctly rather than just matching keywords.
See also: NLP (natural language processing), entity recognition
TechnologyLinkIVR (interactive voice response)
IVR is the classic "press 1 for... press 2 for..." technology in phone switchboards. An AI phone agent often replaces IVR because the caller can simply say what they want instead of navigating a keypad menu.
See also: call flow, phone switchboard
TelephonyLink
K
knowledge base
A knowledge base is the collection of information (opening hours, services, prices, common questions) that an AI agent draws its answers from. The more complete and up to date the knowledge base is, the more accurate the answers.
See also: RAG (retrieval-augmented generation), AI agent
TechnologyLink
L
landing page
A landing page is a web page built for a specific purpose, for example converting visitors from an ad campaign, often with a focused message and a clear CTA.
See also: call to action (CTA), conversion rate
MarketingLinklead capture
Lead capture is the process of collecting contact details and interest from a potential customer, for example through a chat conversation, so the sales team can follow up. One of an AI agent's core jobs beyond answering questions.
See also: lead scoring, marketing qualified lead (MQL)
MarketingLinklead nurturing
Lead nurturing is building a relationship with potential customers over time (through follow-up emails, relevant content or reminders) until they're ready to buy.
See also: lead scoring, funnel model
MarketingLinklead scoring
Lead scoring is ranking captured leads by how sales-ready they appear to be, often based on the questions they asked or the information they provided. Helps the sales team prioritise which leads to contact first.
See also: lead capture, sales qualified lead (SQL)
MarketingLinklive handoff
Live handoff is when an AI conversation is transferred to a human staff member, for example when the question is too complex or the visitor explicitly asks to speak to a person. An important safety valve so the AI doesn't answer incorrectly on sensitive matters.
See also: escalation, fallback response
ProductLinkLLM (large language model)
An LLM (large language model) is an AI model trained on large amounts of text to understand and generate language. Aveexia combines LLMs with a knowledge base (RAG) to answer accurately based on your specific company's information, not just general knowledge.
See also: RAG (retrieval-augmented generation), prompt
TechnologyLinklow-code
Low-code is similar to no-code but allows some custom code for more advanced needs, sitting between off-the-shelf software and building everything from scratch.
See also: no-code, API integration
TechnologyLink
M
marketing qualified lead (MQL)
An MQL is a lead that has shown enough interest (for example by downloading material or asking specific questions) that the marketing team considers it ready to be passed to the sales team.
See also: sales qualified lead (SQL), lead scoring
MarketingLinkmissed call
A missed call is an incoming phone call that no one manages to answer. A common source of lost customers for businesses without an AI phone agent or sufficient staffing. Every missed call is a potential deal that goes to a competitor instead.
See also: queue management, phone switchboard
TelephonyLinkmulti-tenant
Multi-tenant architecture means a single software instance serves multiple different customers at once, with their data kept separate and secure. Common in SaaS products to keep the cost per customer down.
See also: cloud service (SaaS), scalability
TechnologyLinkmulti-turn conversation
A multi-turn conversation is a dialogue where the AI keeps track of earlier messages in the same session, so the visitor can ask follow-up questions without repeating themselves. The opposite is a conversation where each question is answered in isolation.
See also: context window, session
TechnologyLink
N
Net Promoter Score (NPS)
NPS measures how likely a customer is to recommend a business to others, on a 0–10 scale. A broader loyalty metric than CSAT, which measures a single interaction.
See also: customer satisfaction (CSAT), churn
MeasurementLinkNLP (natural language processing)
NLP, or natural language processing, is the AI field concerned with getting computers to understand, interpret and generate human language. It is the foundation that lets an AI agent understand questions phrased in everyday words instead of exact commands.
See also: conversational AI, intent recognition
TechnologyLinkno-code
No-code refers to tools and platforms that let users build or customise functionality without writing code, often through drag-and-drop or filling in forms.
See also: low-code, Zapier integration
TechnologyLinknotice period
The notice period is how far in advance a party must announce they want to end an agreement before it terminates. Important to check in a SaaS agreement before committing.
See also: SaaS agreement, subscription model
BusinessLink
O
omnichannel
Omni-channel means a customer can reach a business through multiple channels (web chat, phone, SMS, WhatsApp) with the conversation staying connected regardless of which channel is used. Delivers a coherent customer experience instead of separate, isolated channels.
See also: web-based chat, WhatsApp Business API
MarketingLinkopening-hours widget
An opening-hours widget is a small component on the website showing current opening hours, often connected to the same information the AI agent uses to answer questions about when the business is open.
See also: knowledge base, chat widget
ProductLink
P
phone switchboard
A phone switchboard routes incoming calls to the right department or person within a company. An AI phone switchboard does the same but with an AI that understands natural speech instead of keypad choices, and can also answer questions itself without transferring the call.
See also: AI phone agent, call flow
TelephonyLinkprompt
A prompt is the instruction or question sent to an AI model to get a desired response. For an AI agent, the prompt typically includes both the visitor's question and relevant information from the knowledge base.
See also: prompt engineering, LLM (large language model)
TechnologyLinkprompt engineering
Prompt engineering is the craft of formulating instructions to an AI model so it gives consistent, accurate and appropriately worded responses. It shapes, among other things, an AI agent's tone of voice and how it handles sensitive or unclear questions.
See also: prompt, conversation design
TechnologyLink
Q
queue management
Queue management describes how a system handles calls or messages that arrive faster than a team can answer. AI can handle recurring questions, but capacity and response times depend on the service and its limits.
See also: missed call, phone switchboard
TelephonyLink
R
RAG (retrieval-augmented generation)
RAG is a technique where an AI first searches a knowledge base for relevant facts and then uses those facts to formulate its answer, instead of guessing based on what it learned during training. This lets Aveexia answer based on your company's actual information rather than general assumptions.
See also: knowledge base, LLM (large language model)
TechnologyLinkrole-based access
Role-based access means different users in a system only get access to the information and functions relevant to their role, for example a support agent not being able to change billing settings.
See also: two-factor authentication, audit log
SecurityLinkRPA (robotic process automation)
RPA is software robots that perform repetitive, rule-based digital tasks, for example copying data between systems. Differs from conversational AI in that RPA doesn't hold dialogues, only executes predefined steps.
See also: hyperautomation, API integration
TechnologyLink
S
SaaS agreement
A SaaS agreement sets out the terms for using a cloud service (price, notice period, data protection, SLA) and is the contract a business signs when it starts using, for example, Aveexia.
See also: notice period, data processing agreement (DPA)
BusinessLinksales qualified lead (SQL)
An SQL is a lead the sales team has assessed as ready for a direct sales conversation, one step further in the funnel than an MQL. An AI agent can help qualify leads into SQLs through its conversations.
See also: marketing qualified lead (MQL), lead scoring
MarketingLinksandbox environment
A sandbox environment is a separate test environment some providers offer for trying changes without affecting the live service.
See also: test environment, AI onboarding
TechnologyLinkscalability
Scalability is a system's ability to handle more users or load without performance degrading. Actual capacity depends on the service plan and operation.
See also: multi-tenant, uptime/SLA
TechnologyLinkself-service
Self-service lets customers find answers or complete simple tasks without contacting staff. AI can assist with those requests, while exceptions are passed to the team.
MarketingLinksentiment analysis
Sentiment analysis is AI technology that assesses the emotional tone of a message, for example whether a customer is frustrated, satisfied or neutral. Can be used to automatically escalate frustrated customers to a human faster.
See also: escalation, live handoff
TechnologyLinksession
A session is a continuous conversation between a visitor and the AI agent, from the first message until the conversation ends or times out. Data within a session is kept together so the AI remembers the context.
See also: multi-turn conversation, context window
TechnologyLinkSMS automation
SMS automation is automatic text messages sent without manual involvement, for example booking confirmations or reminders that an AI agent triggers after an appointment is booked.
See also: automatic response, booking system
ProductLinkspeech synthesis
Speech synthesis is the technical process of generating artificial speech from text, a synonym for text-to-speech (TTS), that makes an AI phone agent sound natural when talking to the caller.
See also: text-to-speech (TTS), voice AI
TechnologyLinkspeech-to-text (STT)
STT (speech-to-text) is the technology that converts spoken language into written text, so an AI can interpret and respond to what someone says on the phone. The reverse of TTS.
See also: text-to-speech (TTS), voice AI
TechnologyLinksubscription model
A subscription model means the customer pays a recurring fee, usually monthly, for continued access to a service, rather than a one-time cost.
See also: SaaS agreement, notice period
BusinessLink
T
test environment
A test environment is used to try changes before launch. Available test environments depend on the provider.
See also: sandbox environment, A/B testing
TechnologyLinktext-to-speech (TTS)
TTS (text-to-speech) is the technology that converts written text into spoken, natural-sounding speech. Used by AI phone agents to read answers aloud to the caller.
See also: speech-to-text (STT), voice AI
TechnologyLinktraining data
Training data is the text and information an AI model learns from during its original training. Different from the knowledge base, which is the company-specific information the AI retrieves answers from for each individual question.
See also: fine-tuning, knowledge base
TechnologyLinktranscription
Transcription is converting spoken content into written text, for example to keep a searchable log of phone calls handled by an AI phone agent.
See also: speech-to-text (STT), AI summary
TechnologyLinktrigger words
Trigger words are specific words or phrases that cause an AI or automated system to perform a particular action, for example escalating to a human when the word "complaint" is mentioned.
See also: escalation, automatic response
TechnologyLinktwo-factor authentication
Two-factor authentication (2FA) requires two separate ways of proving your identity when logging in, for example a password plus a one-time code, to protect accounts from unauthorised access.
See also: encryption, role-based access
SecurityLink
U
uptime/SLA
Uptime is the share of time a service is actually available and working; an SLA (service level agreement) is the vendor's written promise of what uptime level you can expect. Important for an AI agent meant to be available outside regular business hours.
See also: cloud service (SaaS), scalability
TechnologyLink
V
virtual assistant
A virtual assistant is a broader term for AI that performs tasks or answers questions on behalf of a user, in text or speech. An AI agent is a type of virtual assistant specialised in customer interactions on a website.
See also: voice assistant, digital worker
ProductLinkvoice AI
Voice AI is an umbrella term for AI technology that handles speech. Converting speech to text (STT), generating natural-sounding speech (TTS) and interpreting the meaning of what's said. The foundation for every AI phone agent and voicebot.
See also: text-to-speech (TTS), speech-to-text (STT)
TechnologyLinkvoice assistant
A voice assistant is an AI that interacts through speech, such as an AI phone agent or a smart speaker. Differs from a text-based chatbot in the channel, not necessarily in the underlying AI technology.
See also: voicebot, virtual assistant
ProductLinkvoicebot
A voicebot is an AI that holds conversations through speech rather than text, for example over the phone or through a smart speaker. Aveexia's AI phone agent is an example of a voicebot.
See also: AI phone agent, speech synthesis
Product / categoryLink
W
web-based chat
Web-based chat refers to chat conversations that happen directly in the browser through a chat widget on the website, as opposed to chat via apps like WhatsApp or SMS.
See also: chat widget, omnichannel
ProductLinkwebhook
A webhook is a way for one system to automatically send information to another system the moment something happens, for example a new booking being pushed straight into your CRM. Works as an automatic, real-time notification between systems.
See also: API integration, Zapier integration
TechnologyLinkWhatsApp Business API
The WhatsApp Business API is the interface that lets businesses connect an AI agent or chatbot to WhatsApp, so customers can ask questions and book appointments in an app they already use daily.
See also: omnichannel, API integration
TechnologyLinkwhite-label
White-label means a service can be sold or presented under another company's own brand, without the original vendor being visible. Relevant for agencies and resellers who want to offer an AI agent under their own name.
See also: cloud service (SaaS), API integration
ProductLink