AI voice and automation glossary

AI voice assistant

An AI voice assistant is software that talks with people by voice, understands what they say and replies in a natural-sounding voice. Businesses use it to answer calls, take messages, book appointments or answer common questions. It can work on its own for simple requests and pass more complicated ones to a person. A receptionist role is just one of many jobs it can be set up to do.

Example: A business can set up an AI voice assistant to answer incoming calls, collect the caller's details and pass urgent requests to a member of staff.

API (API)

An API, short for application programming interface, is a set of agreed rules that lets one piece of software ask another for information or ask it to do something. It works like a waiter between a customer and a kitchen: you make a request in a set format and get a response back. APIs are how different business tools exchange data without anyone copying it by hand.

Example: A voice system can use an API to ask a diary tool which appointment slots are free before offering one to a caller.

Auto attendant

An auto attendant is a phone feature that answers calls with a recorded greeting and lets callers choose where to go, usually by pressing a number or saying a department name. It then sends the call to the right person, team or voicemail. It follows fixed menu options and does not hold a free-flowing conversation, which makes it simpler than a voice agent.

Example: A caller hears a greeting, presses 2 for accounts and is sent straight to the accounts team's phone.

IVR vs AI Voice Assistant: Keep, Replace or Layer?

Automatic speech recognition (ASR)

Automatic speech recognition is the technology that listens to spoken audio and works out which words were said. It is the first step in most voice AI systems, because a computer must turn speech into written words before it can act on them. The term is often used interchangeably with speech to text, although ASR describes the wider field of research and tools.

Example: When a caller says their postcode, automatic speech recognition turns the sound into written characters the system can use.

Barge-in

Barge-in is the ability of a caller to interrupt a voice system while it is speaking, and for the system to stop talking and listen. Without it, callers must wait for long prompts to finish, which feels frustrating. Good barge-in handling makes a conversation feel more like talking to a person, who naturally stops when you cut in.

Example: If the assistant starts reading out opening hours and the caller says they only want to book, barge-in lets it stop and listen.

AI Voice Assistant Myths, and How to Test Each One Yourself

Call recording

Call recording is the saving of the audio of a phone conversation so it can be listened to later. Businesses use it for training, checking quality, settling disputes and keeping a record of what was agreed. Because recordings contain personal data and voices, rules on telling callers and on how long to keep recordings can apply, so check the requirements that cover your business.

Example: A manager listens to a saved call recording to coach a team member on how they handled a difficult conversation.

Call routing

Call routing is the process of deciding where an incoming call should go, such as to a particular person, team, phone number or voicemail. The decision can depend on the time of day, the number dialled, the caller's menu choice or what they say. Good routing means callers reach someone who can help without being passed around.

Example: Calls made outside opening hours are routed to voicemail, while calls during the day go to the next free team member.

IVR vs AI Voice Assistant: Keep, Replace or Layer?

Call whisper

A call whisper is a short message that only the person receiving a call can hear, played just before they are connected to the caller. It tells them useful details, such as which phone number was dialled or what the caller wants. This helps staff answer in the right way, especially when one team handles calls for several services.

Example: Before a call connects, the team member hears a brief whisper saying the caller is asking about a delivery.

Conversational AI

Conversational AI is the broad name for technology that lets people talk with computers in everyday language, by typing or speaking. It combines ways of understanding what a person means with ways of replying sensibly. Chat assistants, voicebots and voice agents are all examples, and it differs from older systems that only accept fixed menu choices or exact keywords.

Example: A customer asks a question in their own words and the conversational AI system answers without making them pick from a menu.

CRM (CRM)

A CRM, or customer relationship management system, is software that stores information about your customers and prospects in one place. It typically records contact details, past conversations, notes and where each person is in your sales process. It helps a team stay organised, avoid forgetting follow-ups and see the history of a relationship without searching through emails or spreadsheets.

Example: After a call, the caller's details and a short note are saved to the CRM so any colleague can follow up.

Data controller

A data controller is the person or organisation that decides why and how personal data is used. Under UK GDPR and GDPR wording, the controller carries the main responsibility for handling that data lawfully. A business that collects customer details for its own purposes is usually the controller. This is a general explanation only and is not legal advice.

Example: A company that decides to collect callers' names to arrange appointments is acting as the data controller for those details.

Data processor

A data processor is a person or organisation that handles personal data on behalf of a data controller and follows the controller's instructions. It does not decide for itself why the data is used. Under UK GDPR and GDPR wording, the relationship is normally set out in a written agreement. Cloud tools and service providers often act as processors.

Example: A service provider that stores a business's call records only on that business's instructions is acting as a data processor.

Is It Legal to Use AI to Answer Business Calls in the UK?

Deflection

Deflection is when a customer's question is resolved through self-service or automation instead of reaching a human agent. Examples include a help article, a chatbot answer or a voice assistant handling a simple request. The aim is to free staff for harder issues, but it only works well if the customer actually gets a useful answer and can still reach a person.

Example: A customer asks about delivery times, gets the answer from an automated assistant and never needs to speak to a member of staff.

Endpointing

Endpointing is how a voice system decides that a person has finished speaking and it is time to respond. It usually listens for a pause of a certain length. If it decides too early, it cuts people off mid-sentence; if too late, the conversation feels slow and awkward. Getting the balance right is an important part of natural-sounding calls.

Example: When a caller pauses after giving their name, endpointing helps the system judge whether they have finished or are still thinking.

First response time

First response time is the length of time between a customer getting in touch and receiving the first reply from your business. The reply can come from a person or from automation. It is a common measure of how responsive a support team is, because customers often judge a business by how quickly they first hear back, even before their problem is solved.

Example: A shop tracks first response time to see how long customers wait before anyone acknowledges their email.

Human handover

Human handover, also called a warm transfer, is when an automated system passes a conversation to a person, ideally sharing what has already been said so the caller does not have to repeat it. It is used when a request is complex, sensitive or urgent, or when the caller asks for a person. A warm transfer means the person is briefed before taking over.

Example: The assistant collects the caller's name and issue, then transfers the call to a team member with a short summary.

Integration

An integration is a connection that lets two or more software tools share information and work together. For example, linking your phone system to your CRM means call details can be saved automatically. Integrations may use an API, a webhook or a built-in connector, and they reduce the need to type the same information into several places.

Example: An integration between a voice system and a calendar lets bookings made on a call appear in the team's diary.

Intent detection

Intent detection is the process of working out what a person wants from what they say or write, such as booking an appointment, asking about a price or making a complaint. It looks at the meaning behind the words rather than exact phrases. The result lets a system choose the right next step, like answering, collecting details or passing the person to a colleague.

Example: When a caller says they need to move Tuesday's visit, intent detection recognises that they want to reschedule.

IVR (IVR)

IVR stands for interactive voice response. It is a phone system that greets callers with recorded options and responds to key presses or simple spoken words, such as press 1 for sales. It can route calls or give basic information. Traditional IVR follows a fixed script, whereas a voice agent can handle open-ended speech.

Example: A caller presses 3 on an IVR menu to hear opening hours without speaking to anyone.

Large language model (LLM)

A large language model is a type of AI trained on very large amounts of text so that it can understand and produce human-like language. It can answer questions, summarise, draft messages and follow instructions written in everyday words. In voice systems it often acts as the part that decides what to say next. Its answers can sometimes be wrong, so checks matter.

Example: A voice system uses a large language model to write a sensible reply to a caller's question in plain language.

Latency (voice)

In voice systems, latency is the delay between a person finishing speaking and hearing the reply. It builds up from several steps, including recognising speech, working out a response and speaking it aloud. Low latency makes a conversation feel natural, while a long pause can leave callers unsure whether the system heard them or has stopped working.

Example: If there is a long silence after the caller asks a question, the system is suffering from high latency.

Why an AI Voice Assistant Pauses Before It Replies

Lead nurturing sequence

A lead nurturing sequence is a planned series of messages sent over time to someone who has shown interest but is not ready to buy. The messages might be emails, texts or calls that share helpful information and stay in touch. The aim is to build trust and keep your business in mind until the person is ready to decide, and it can be automated.

Example: After someone requests a quote, they receive a short series of helpful follow-up emails spread over several weeks.

Lead qualification

Lead qualification is the process of checking whether a new enquiry is likely to become a customer. It usually involves asking a few questions about needs, timing, budget or location, and comparing the answers with what your business can offer. It helps a sales team spend time on the enquiries that fit best, and a voice assistant can ask the first questions.

Example: A caller is asked what service they need and when, so the team can tell whether they are a good fit.

Lead scoring

Lead scoring is a method of giving each potential customer a score to show how likely or ready they are to buy. Points are usually added for useful signals, such as answering key questions or replying to messages. The scoring rules are chosen by the business. Teams use scores to decide whom to contact first.

Example: A prospect who asks for a quote and gives a start date is given a higher score than someone who only browsed.

Missed call text back

Missed call text back is an automatic feature that sends a text message to a caller when their call goes unanswered. The message typically apologises, says the business will be in touch and may ask how to help. It lets the person know they have not been ignored and can start a conversation by message instead of giving up or phoning a competitor.

Example: When nobody can pick up, the caller automatically receives a text asking what they need and when to call them back.

Natural language understanding (NLU)

Natural language understanding is the part of an AI system that works out the meaning of words people say or type. It deals with different ways of saying the same thing, picks out details such as dates or names and helps identify what the person wants. It comes after speech has been turned into text and before the system decides how to respond.

Example: Whether a caller says next Friday afternoon or Friday after lunch, natural language understanding picks out the day and rough time.

No-code automation

No-code automation means building automated workflows using visual tools, such as dragging blocks and filling in forms, instead of writing computer code. It lets business owners and non-programmers connect apps and set up repeated tasks themselves. It suits many everyday jobs, though very complex or unusual needs may still call for a developer.

Example: An office manager sets up a rule so new enquiry forms are copied into a spreadsheet, without writing any code.

No-Code Tools or Custom Automation? How to Choose

RPA (RPA)

RPA, or robotic process automation, is software that copies the clicks and keystrokes a person would make on a computer to complete repetitive tasks. Despite the name, there is no physical robot. It is often used with older systems that have no easy way to connect, such as copying data between screens, and it follows fixed rules rather than making judgments.

Example: A software robot opens an invoice, copies the total into an accounts screen and moves on to the next one.

Rule-based chatbot

A rule-based chatbot is a chat tool that follows a fixed script of options and keywords set in advance. It can only answer what its creators planned for, so it works for simple, predictable questions but struggles when people phrase things differently. It differs from conversational AI, which tries to understand free-form language.

Example: A chatbot offers three buttons for delivery, returns or opening hours and replies with a prewritten answer to each.

Sales pipeline stage

A sales pipeline stage is one step in the journey a potential customer takes from first contact to becoming a customer or being lost. Typical stages might be new enquiry, contacted, quote sent and won, though each business names its own. Tracking stages in a CRM shows where deals are waiting and helps teams decide what to do next.

Example: After sending a quote, the team moves the contact from the contacted stage to the quote sent stage.

Special category data

Special category data is a type of personal data that data protection law treats as more sensitive and gives extra protection. Under UK GDPR and GDPR wording, it includes information about health, racial or ethnic origin, religious beliefs, sexual orientation and genetic or biometric data used to identify someone. Extra conditions apply to using it, so it should be handled with particular care.

Example: A caller mentioning a medical condition while booking an appointment is giving information that may count as special category data.

Speech to text (STT)

Speech to text is technology that converts spoken words into written text. In a voice system, it lets the software read what a caller said so it can understand and act on it. It is also used to create call transcripts and meeting notes. Accuracy can be affected by accents, background noise and call quality.

Example: A caller's spoken request is turned into text so the system can read it and decide what to do.

Text to speech (TTS)

Text to speech is technology that turns written words into spoken audio. In a voice system, it is how the computer talks back to the caller after it has decided what to say. Voices can vary in language, accent and tone, and modern ones are designed to sound natural rather than robotic, though quality differs between providers.

Example: The system writes a reply in text and text to speech reads it aloud to the caller.

Ticket triage

Ticket triage is the process of sorting incoming support requests, called tickets, so that each is given the right category, priority and owner. Urgent or serious issues go to the front of the queue, and simple ones can be answered quickly or automated. Good triage means customers reach the right person sooner and nothing important gets buried.

Example: A message saying a service is completely down is marked urgent and sent to a senior team member first.

Trigger (automation)

A trigger is the event that starts an automated workflow. It might be a form being submitted, an email arriving, a call ending or a set time of day. Once the trigger happens, the automation carries out its steps without anyone starting it by hand. Every automated workflow needs a trigger to know when to begin.

Example: When a new enquiry form is submitted, that submission is the trigger that starts the follow-up workflow.

Voice agent

A voice agent is an AI system that holds spoken conversations with people and can take actions, such as booking an appointment, updating a record or passing the call to a colleague. It listens, works out what is needed, replies aloud and connects to other business tools. It goes further than a simple menu because callers can speak naturally.

Example: A voice agent answers a call, books a visit in the diary and sends the caller a confirmation message.

Voicebot

A voicebot is an automated program that talks with callers by voice. Some voicebots follow a fixed script and a limited set of phrases, while others use conversational AI to cope with free-form speech. The word is a broad label, and what a particular voicebot can do depends on how it is built, so it is worth checking its actual abilities.

Example: A voicebot answers the phone, asks how it can help and gives the caller basic information.

Website Chatbot or Phone Assistant? Choosing for a Small Business

Voicemail

Voicemail is a phone service that lets callers leave a spoken message when nobody answers. The message is stored so the person can listen later, and some systems turn it into text or send it by email. It is a simple way to catch missed calls, though callers may hang up rather than leave a message, so quick follow-up is important.

Example: A caller leaves a voicemail after hours, and the team listens to it the next morning.

Webhook

A webhook is an automatic message that one piece of software sends to another as soon as something happens, such as a call ending or a form being submitted. The receiving tool can then act straight away. Unlike an API request, where you keep asking whether anything is new, a webhook tells you the moment there is news.

Example: When a call finishes, the phone system sends a webhook so the call summary is saved without any manual step.

Triggers and Actions Explained: How Automations Actually Work

Workflow automation

Workflow automation is the use of software to carry out a series of business tasks automatically, following rules you set. When a trigger happens, such as a new enquiry, the software performs the steps in order, like saving details, sending a message and alerting a colleague. It reduces repetitive manual work and helps prevent steps being forgotten.

Example: A new enquiry is saved to the CRM, the customer receives a confirmation and a team member is notified, all automatically.

GoHighLevel vs HubSpot for Small Businesses: Which Should You Automate On?