Glossary

Plain definitions of the words used around AI receptionists, call handling and the systems they connect to. Each entry stands on its own.

Categories of service

After-hours answering: After-hours answering is the handling of a business’s calls outside its normal opening hours: evenings, nights, weekends and holidays. It is done by a person or system other than the in-house team.
AI receptionist: An AI receptionist is software that answers a business’s phone calls and holds a spoken conversation with the caller. It greets them, answers questions, takes details, books appointments and routes calls, with no person on the line.
Answering service: An answering service is a business that answers phone calls on behalf of other businesses. It typically takes messages, relays urgent calls and gives basic information when the client cannot answer.
Bilingual answering: Bilingual answering is call handling in which the receptionist or service can hold the conversation in two languages and switch to the caller’s language when needed. In the United States that is most often English and Spanish.
Call center: A call center is a department or company that handles a high volume of phone calls for an organization, using teams of agents and specialized routing software. Calls may be inbound (customer service, sales, support) or outbound (sales, collections, surveys).
Front desk: The front desk is the reception function of a business: the person or system that greets visitors and callers. It answers first-line questions, takes details and directs each inquiry to the right place.
HIPAA-compliant answering: HIPAA-compliant answering is call handling for US healthcare providers in which the answering provider operates as a business associate under HIPAA. That means a signed business associate agreement and safeguards for any protected health information callers disclose.
Live answering: Live answering is a call being picked up and handled in real time, by a person or a conversational system. The alternative is voicemail, an automated menu or a recorded message.
Overflow answering: Overflow answering is the handling of calls a business cannot answer itself during working hours, because every line is busy or nobody picks up within a set number of rings. Those calls forward to a backup service.
Virtual receptionist: A virtual receptionist is a receptionist who answers a business’s calls from somewhere other than the business’s own premises. Usually they work for a receptionist service shared across several client businesses.

Telephony and call handling

Abandonment rate: Abandonment rate is the percentage of incoming calls in which the caller hangs up before being answered, usually while ringing or waiting in a queue.
ACD: An ACD (automatic call distributor) is the part of a phone system that receives incoming calls and routes each one to a specific agent or group. It follows rules such as skills, availability, priority or order of arrival.
Answer rate: Answer rate is the percentage of incoming calls that are answered by a person or system, as opposed to going unanswered, being abandoned, or reaching voicemail.
Barge: In telephony, barge (or call barge) is a supervisor joining a live call between an agent and a caller so that all three parties can hear and speak.
Call overflow: Call overflow is the routing of incoming calls to a secondary destination (another team, an outside answering service, or an automated system) when the primary destination is busy, unstaffed or not answering within a set time.
Call queue: A call queue is the holding pattern a phone system places incoming calls in when no one is available to answer. Calls are released in order, usually first in, first out, as agents become free.
Cold transfer: A cold transfer, also called a blind transfer, is a call transfer in which the call is passed straight to the recipient without any introduction. The recipient answers with no context.
Concurrency: Concurrency, in telephony, is the number of calls a line, phone system or service can carry at the same time. A call that arrives when every channel is in use gets a busy signal, a queue or an overflow destination.
DID: A DID (direct inward dialing number) is a phone number that routes straight to a specific destination, such as a person, a department or an automated receptionist, without passing through a switchboard.
DNI: DNI (dynamic number insertion) is a call-tracking technique in which a website shows a different phone number to each visitor depending on how they arrived: the ad, campaign or search that brought them. Calls can then be attributed to their source.
DTMF: DTMF (dual-tone multi-frequency) is the signaling system behind touch-tone dialing. Each key on a phone keypad sends a distinct pair of audio tones that equipment on the line can recognize.
First call resolution: First call resolution (FCR) is the share of calls in which the caller’s need is fully handled during that call, with no transfer, callback or repeat contact required.
Inbound call: An inbound call is a call placed to a business’s number by a caller, as opposed to an outbound call the business places. Answering inbound calls is the core job of a receptionist, human or AI.
IVR: An IVR (interactive voice response) system is an automated phone menu that plays recorded prompts and routes callers based on the keys they press or the words they say.
Monitor: Monitor, or silent monitoring, is a supervision feature that lets a supervisor listen to a live call without either the agent or the caller hearing them.
Number porting: Number porting is moving an existing phone number from one carrier or service provider to another, so the business keeps the same number when it changes provider.
PSTN: The PSTN (public switched telephone network) is the worldwide, interconnected network of public telephone systems (landline, mobile and the carriers between them) that a phone number is reachable on.
Screen pop: A screen pop is the automatic display of a caller’s record (name, history, open orders or tickets) on the answering agent’s screen as the call arrives, matched on the caller’s phone number.
SIP: SIP (Session Initiation Protocol) is the standard signaling protocol used to set up, modify and end voice and video calls over IP networks. A SIP trunk is a SIP connection from a phone system to a carrier.
SMS: SMS (short message service) is the standard for sending short text messages between phones over the carrier network. In a receptionist context it is the channel for texting a caller a link, a confirmation or a follow-up.
Speed to answer: Speed to answer is the time between a call arriving and it being picked up. It is reported as an average (average speed of answer, ASA) or as the share of calls answered within a target such as twenty seconds.
Voicemail: Voicemail is a system that records a spoken message from a caller when the called party does not answer, storing it for later playback or delivery by email or text.
VoIP: VoIP (voice over Internet Protocol) is the technology that carries phone calls as data packets over an IP network such as the internet instead of over dedicated circuit-switched phone lines.
Warm transfer: A warm transfer is a call transfer in which the person handing off the call first speaks to the recipient, explains who is calling and why, and only then connects the caller.
WebRTC: WebRTC (Web Real-Time Communication) is an open standard built into modern browsers. It lets a web page capture a microphone or camera and exchange live audio and video with another party without a plug-in or separate app.
Whisper: Whisper is a call-handling feature that lets a supervisor speak privately to the agent on a live call without the caller hearing. The same word covers a short announcement played to the answering party before a call is connected.

How voice AI works

AI agent: An AI agent is a system built around a language model that pursues a goal by deciding on and carrying out a sequence of actions: asking questions, calling tools, checking results. It does more than produce a single reply to a single prompt.
Barge-in: In speech AI, barge-in is the caller speaking while the system is still talking, and the system’s ability to detect that, stop, and respond to what was said.
Context window: The context window is the maximum amount of text (instructions, retrieved information and the conversation so far) that a language model can take into account when producing its next reply. It is measured in tokens.
Conversational AI: Conversational AI is software that holds a natural back-and-forth exchange with a person in speech or text. It understands free-form input and replies in kind rather than presenting a fixed menu.
Diarization: Diarization is dividing an audio recording by speaker, working out who spoke when, so that a transcript can attribute each passage to the right person.
Endpointing: Endpointing is detecting where a spoken utterance begins and ends in a stream of audio, so that speech recognition and the reply are triggered at the right moment.
Function calling / tool use: Function calling, or tool use, is a language model’s ability to request that a defined external operation be run (look up a record, check a calendar, create a booking), receive the result, and continue the conversation using it.
Grounding: Grounding is the practice of tying a language model’s answers to a specific, trusted source of information (documents, records or live system data). The model answers from what is provided rather than from what it can generate.
Guardrails: Guardrails are the rules and checks placed around an AI agent to keep its behavior inside agreed limits: what it may discuss, what it must never say or do, and when it must hand off to a person.
Hallucination: A hallucination is a fluent, confident statement by a language model that is false or unsupported: an invented fact, price, policy or appointment slot presented as if it were known.
Intent detection: Intent detection is the step in which a conversational system works out what the caller is trying to achieve (book, cancel, ask about an order, reach a person) from the words they used.
Interruption handling: Interruption handling is how a voice agent behaves when the caller speaks while it is still talking: whether it notices, how quickly it stops, and what it does with what the caller said.
Large language model (LLM): A large language model (LLM) is a neural network trained on very large amounts of text to predict likely continuations. That lets it understand instructions, hold conversations and generate fluent language.
Latency (time to first token): In voice AI, latency is the delay between the caller finishing an utterance and the agent starting to speak in reply. Time to first token (TTFT) is the language-model portion of that delay, measured to the first word generated.
Natural language understanding (NLU): Natural language understanding (NLU) is the part of a conversational system that extracts meaning from free-form text or transcribed speech: intents, entities such as names, dates and addresses, and the relationships between them.
Prompt injection: Prompt injection is an attack in which text supplied to a language model (by a caller, a document or another system) is crafted to be read as instructions, overriding or subverting the rules the model was given.
Retrieval-augmented generation (RAG): Retrieval-augmented generation (RAG) is a technique in which relevant passages are fetched from a document store at question time and given to a language model. Its answer is then based on those documents rather than on memory alone.
Sentiment analysis: Sentiment analysis is the automatic assessment of the emotional tone of speech or text: positive, neutral or negative, and often finer states such as frustration or urgency. It reads the words used and, in voice, how they were said.
Speech-to-text (STT): Speech-to-text (STT), also called automatic speech recognition (ASR), is the conversion of spoken audio into written words so that software can process what was said.
System prompt: A system prompt is the standing set of instructions given to a language model before a conversation starts. It defines who the model is, what it may and may not do, and how it should behave.
Text-to-speech (TTS): Text-to-speech (TTS) is the generation of spoken audio from written text, producing the voice a caller hears when an AI agent replies.
Turn detection: Turn detection is how a voice system decides that the caller has finished speaking and it is now its turn to respond. It also covers the reverse: noticing the caller has started speaking and stopping.
Turn-based context awareness: Turn-based context awareness is a conversational system’s ability to carry state across the turns of a conversation. A later utterance is understood in light of earlier ones, and nothing already answered is asked again.
Voice activity detection: Voice activity detection (VAD) is the technique of deciding, moment by moment, whether an audio signal contains human speech as opposed to silence or background noise.
Voice agent: A voice agent is a conversational AI that talks with people over a voice channel (a phone line or a browser), listening, deciding and speaking in real time. It often takes actions such as booking or looking something up.
Voice cloning: Voice cloning is the creation of a synthetic text-to-speech voice that imitates a specific real person, built from recordings of their speech.
Word error rate: Word error rate (WER) is the standard measure of speech-recognition accuracy. It is the number of substituted, deleted and inserted words in a transcript, divided by the number of words in the correct reference.

Integrations

API: An API (application programming interface) is a defined set of requests one piece of software can make to another to read or change its data.
Calendar booking: Calendar booking is the ability to check availability in a calendar and reserve a slot for a caller during the conversation.
CRM sync: CRM sync keeps call outcomes (contact details, notes, summaries, follow-up tasks) in step with a customer relationship management system, so records are created or updated without manual entry.
Custom call flow: A custom call flow is the designed sequence of steps a receptionist follows on a call, tailored to one business: what it asks, what it looks up, where it routes, and what happens afterwards.
Custom integration: A custom integration is a connection between a receptionist and a business’s own systems that the provider builds and maintains, rather than one the business assembles from a template, a webhook or an automation platform.
Post-call vs real-time integration: A post-call integration pushes a transcript, summary or record to another system after a call ends. A real-time integration lets the receptionist query a system during the call and answer the caller from live data.
Webhook: A webhook is an HTTP request that one system sends to a URL you provide when an event happens, carrying the event data as its payload.
Zapier: Zapier is a third-party automation platform that connects thousands of web applications through trigger-and-action workflows called Zaps, without writing code.

Receptionist features

Agent Persona: An agent persona is the defined identity of an AI receptionist: its name, voice, tone, manner and goals. It shapes how the receptionist speaks and behaves on every call.
Barge, steer and join: Barge, steer and join are three levels of live intervention in an AI-handled call: watching it in real time, giving the AI instructions mid-call, and joining the call as a person.
Built-in CRM: A built-in CRM is a contact and interaction database included inside a product. Call history, contact details and tags are kept without connecting an external customer relationship management system.
Call feedback: Call feedback is the mechanism by which a business reviews an individual call and tells the receptionist what it should have done differently. Future calls improve as a result.
Call recording: Call recording is the capture and storage of the audio of a phone call so it can be replayed later for review, training, dispute resolution or compliance.
Call sentiment: Call sentiment is an automated estimate of how a caller felt during a call (positive, neutral or negative), derived from their words and sometimes their tone.
Call transcription: Call transcription converts the audio of a phone call into written text, either live during the call or from a recording afterwards.
Caller memory: Caller memory is the ability to recognize a returning caller, usually by phone number, and carry context from previous conversations into the current one.
Co-browse: Co-browse is a session in which an agent, human or AI, sees or guides the page a visitor is on. It can walk them through a form, a checkout or a booking while talking to them.
Customer portal: A customer portal is the signed-in web application where a business manages its receptionist (settings, call flows, contacts, recordings, transcripts, billing) and reviews what happened on its calls.
Embedded receptionist: An embedded receptionist is the same AI receptionist placed inside a business’s own website or portal through a JavaScript integration, so visitors can talk to it without dialing a number.
Message delivery: Message delivery is the route by which a call’s summary and details reach the business after the call: an email inbox, a text, a chat channel, an automation platform or a webhook.
Omnichannel: Omnichannel means a receptionist handles conversations across several channels (phone, in-browser voice, SMS, email, chat, a customer portal, webhooks) with the same knowledge and the same record of the customer.
Outbound calling: Outbound calling is a receptionist or agent placing calls rather than answering them, for example to confirm an appointment, follow up a lead or return a missed call.
Portal voice assistant: A portal voice assistant is a conversational helper inside a product’s customer portal. It answers questions about the product and walks the user through setup and configuration by voice.
Post-call SMS confirmation: A post-call SMS confirmation is a text message sent to the caller after the call restating what was captured (an appointment, an address, a request), so they can confirm or correct it.
Post-call summary: A post-call summary is a short, structured write-up generated when a call ends: who called, why, what was agreed and what needs doing. It is sent to the business or written to its systems.
Real-time CRM lookup: Real-time CRM lookup is a receptionist reading a customer record from a CRM during the call, so it can greet a known caller and answer from live data instead of taking a message.
Real-time escalation: Real-time escalation is handing a live call from an AI receptionist to a person (the business owner, a colleague, or a human answering desk) the moment a defined condition is met.
Scheduling: Scheduling is the receptionist’s ability to offer available times and book, reschedule or cancel appointments for a caller during the call.
Team management: Team management is the ability to invite several people from a business into a product’s portal and control what each of them can see and change.
Transcripts: Transcripts are the stored, readable text of past calls, kept alongside recordings and summaries so a conversation can be reviewed word for word.
Voice over the internet: Voice over the internet is a spoken conversation carried over a data connection rather than the telephone network, typically from a button on a web page using WebRTC.
Voice selection: Voice selection is choosing the synthetic voice a receptionist speaks with (its gender, accent, age and style) from a set of available text-to-speech voices.
Webhook comms: Webhook comms is the use of webhooks as a communication channel. Call events, messages and outcomes are delivered as HTTP requests to a URL the business controls, for its own systems to act on.
Workflow builder: A workflow builder is a visual editor in which a business assembles what its receptionist does from a set of pre-built nodes, without code: the questions, lookups, branches, transfers and follow-up actions.
Workflow generator: A workflow generator drafts a starting workflow for a receptionist from a plain description of the business, for the owner to edit rather than build from a blank canvas.

Compliance

AI disclosure: AI disclosure is telling a person, clearly and early, that they are interacting with an artificial intelligence rather than a human. A growing number of laws require it for bots and AI voices used in commerce.
BAA: A BAA (business associate agreement) is the written contract HIPAA requires between a covered entity and any vendor that handles protected health information for it. It sets out what the vendor may do with the data and how it must protect and report on it.
Call recording disclosure: A call recording disclosure is the notice given at the start of a call that it is being recorded or monitored, so that the participants can consent by continuing or end the call.
Consent mode: Consent mode is Google’s mechanism for telling its advertising and analytics tags whether a website visitor has consented to cookies. The tags then adjust what they store and send instead of firing regardless.
DNC registry: The National Do Not Call Registry is the US list, run by the Federal Trade Commission, of phone numbers whose owners have asked not to receive telemarketing calls. Sellers and telemarketers must check their calling lists against it.
GDPR: The GDPR (General Data Protection Regulation) is the European Union’s data-protection law, in force since May 2018. It governs how organizations anywhere in the world collect, use and store personal data about people in the EU and EEA.
HIPAA: HIPAA (the Health Insurance Portability and Accountability Act of 1996) is the US federal law whose Privacy, Security and Breach Notification Rules govern how health plans, healthcare providers, clearinghouses and their business associates may use, disclose and protect individually identifiable health information.
PHI: PHI (protected health information) is any individually identifiable health information, about a person’s condition, care or payment for care, that is held or transmitted by a HIPAA covered entity or its business associate, in any form.
Prior express written consent: Prior express written consent is the TCPA’s standard for marketing robocalls and texts. It is a signed written agreement in which the person clearly authorizes a named seller to contact a specific number using an autodialer or an artificial or prerecorded voice.
TCPA: The TCPA (Telephone Consumer Protection Act) is the 1991 US federal law that restricts calls and text messages made with an autodialer or an artificial or prerecorded voice. It requires the recipient’s consent before most such calls are placed.
Two-party consent: Two-party consent, more precisely all-party consent, is a rule under some US state laws that a phone call may be recorded only if everyone on the call has agreed. Under federal law and in most states, the consent of one participant is enough.

Commercial terms

Cost of a missed call: The cost of a missed call is the expected revenue a business forgoes each time an inbound call goes unanswered. It is estimated from the value of a typical customer, the share of calls that are new inquiries, and the chance an unanswered caller does not call back.
Cost per lead: Cost per lead (CPL) is the average amount a business spends on marketing to generate one new inquiry. It is calculated by dividing total marketing spend over a period by the number of leads produced in that period.
Google Local Services Ads (LSA): Google Local Services Ads (LSA) are pay-per-lead advertisements shown at the top of Google search results for local service businesses. The advertiser is charged for each qualifying call or message rather than for each click.
Lead qualification: Lead qualification is asking a new inquirer a set of questions to establish whether, and how urgently, the business can help them. Sales effort then goes to the inquiries most likely to become customers.
Speed to lead: Speed to lead is the time between a prospect first making contact (a call, form, chat or ad click) and the business responding in a way that starts a conversation.

Ready to stop missing calls?

Try it free, or ask the team how it would work for your business.

Talk to the team

Call the AI