Reference
AI Voice Agent Glossary
Clear definitions for every term you'll encounter when evaluating, deploying, or scaling AI voice agents across your organization.
AI Voice Agent
An artificial intelligence system that conducts human-like phone conversations autonomously, without requiring a live human operator. AI voice agents can answer questions, qualify leads, schedule appointments, and escalate complex issues — operating 24/7 at scale across any industry.
Read definition →
AHT (Average Handle Time)
A workforce management metric measuring the average duration of a customer interaction from initial contact to call termination. AHT typically includes talk time plus after-call work (ACW) such as note-taking and follow-up preparation. Lower AHT often correlates with efficiency, but balancing it against quality and customer satisfaction is essential.
Read definition →
Agentic AI
AI systems capable of taking multi-step, goal-directed actions without continuous human intervention. Agentic AI can browse systems, execute workflows, update CRMs, send follow-up messages, and make decisions across multiple tools — going far beyond simple question-and-answer interactions.
Read definition →
API (Application Programming Interface)
A set of defined rules and protocols that allow different software applications to communicate with each other. In AI voice agent contexts, APIs connect the voice platform to CRMs, scheduling tools, databases, and other business systems to exchange data in real time during a call.
Read definition →
ASA (Average Speed of Answer)
A customer service metric that measures the average time between when a call enters a queue and when an agent answers it. ASA is a key performance indicator in contact centers, with targets often set between 20 and 45 seconds. Lower ASA indicates better responsiveness and typically improves customer satisfaction.
Read definition →
ASR (Automatic Speech Recognition)
The technology that converts spoken audio into machine-readable text in real time. Also called Speech-to-Text (STT), ASR is the first stage of every AI voice agent pipeline — its accuracy directly determines how well downstream intent detection and response generation perform.
Read definition →
ACD (Automatic Call Distributor)
A telephony system that automatically routes incoming calls to the appropriate agent or team based on predefined rules, skills, or availability. Modern ACDs form the backbone of contact center infrastructure, queuing calls, managing hold times, and distributing workload to maximize efficiency and customer satisfaction.
Read definition →
ATDS (Automatic Telephone Dialing System)
A technology system that automatically dials a list of phone numbers and delivers pre-recorded messages or connects callers to available agents. ATDS systems are regulated by the TCPA and require explicit consent; they are commonly used for outbound campaigns, appointment reminders, and debt collection.
Read definition →
Abandonment Rate
A contact center metric that measures the percentage of inbound calls that are disconnected or abandoned before an agent answers them. High abandonment rates indicate insufficient staffing or long wait times, directly impacting customer satisfaction and revenue, particularly in inbound sales and customer service environments.
Read definition →
BAA (Business Associate Agreement)
A contractual agreement required under HIPAA between a healthcare organization and any third-party vendor (including AI voice platform providers) that handles Protected Health Information (PHI). The BAA specifies how PHI will be protected, accessed, and ultimately destroyed, ensuring compliance with federal healthcare privacy law.
Read definition →
Barge (Call Barge / Whisper / Monitor)
A call center feature that allows a supervisor or quality manager to listen in on a live call between an agent and a customer — either in whisper mode (hearing both sides but not heard by the customer), monitor mode (listening only), or barge mode (joining the call to coach or assist the agent in real time).
Read definition →
Compliance
Adherence to regulatory frameworks governing how AI systems collect, store, and process data. For AI voice agents, compliance typically covers HIPAA (healthcare), TCPA (telemarketing), GDPR (EU data privacy), and PCI-DSS (payment data) — each imposing rules on call recording, consent, and data retention.
Read definition →
CSAT (Customer Satisfaction Score)
A metric measuring customer satisfaction with a specific interaction or service on a numerical scale, typically 1–5 or 1–10. CSAT surveys are usually administered immediately after a call or transaction. Organizations track CSAT as a key operational metric to identify service gaps and measure the impact of agent training or AI agent improvements.
Read definition →
Conversational AI
A category of artificial intelligence designed to simulate natural human dialogue through text or voice interfaces. Conversational AI combines NLP, intent detection, and response generation to create interactions that feel less like navigating a menu and more like speaking with a knowledgeable person.
Read definition →
CRM Integration
The connection between an AI voice platform and a Customer Relationship Management system such as Salesforce, HubSpot, or Zoho. CRM integration allows the agent to look up caller records, log call outcomes, update contact fields, and trigger follow-up workflows automatically during or after each call.
Read definition →
Call Deflection
A contact center strategy that routes incoming calls away from live agent queues toward self-service channels — including IVR, chatbots, or AI voice agents — to reduce agent workload and operational costs. Effective call deflection improves throughput and reduces average handle time without sacrificing customer satisfaction if the self-service option is high-quality.
Read definition →
CCaaS (Contact Center as a Service)
A cloud-based contact center platform delivered via subscription, eliminating the need for on-premise hardware and infrastructure. CCaaS solutions like Zendesk, Twilio, and Amazon Connect provide call routing, IVR, analytics, and integration APIs, enabling organizations to rapidly deploy and scale customer service operations.
Read definition →
EHR (Electronic Health Record)
A digital version of a patient's complete medical history maintained by a healthcare provider or health system. EHRs consolidate clinical notes, medications, lab results, and imaging studies, allowing clinical staff quick access to patient information during consultations. AI voice agents can integrate with EHR systems to confirm appointments, deliver test results, and triage patient concerns.
Read definition →
EMR (Electronic Medical Record)
A digital record of a patient's medical history maintained by a specific healthcare provider or clinical practice. Unlike EHRs, EMRs are typically limited to one practice and not designed for interoperability across organizations. AI voice agents can integrate with EMR systems to retrieve appointment history, medication information, and previous visit summaries.
Read definition →
Erlang C
A mathematical formula used in contact centers to calculate the number of agents required to handle an expected volume of calls at a target service level. Erlang C accounts for call arrival rate, average handle time (AHT), and acceptable wait time, helping workforce management teams right-size staffing to meet demand without overstaffing.
Read definition →
Escalation
The process of transferring a customer interaction from one channel, agent, or system to another — typically when an issue is too complex or sensitive for the current handler to resolve. AI voice agents recognize escalation triggers and smoothly route callers to a human agent, ensuring critical issues receive appropriate human judgment and empathy.
Read definition →
Hallucination
A phenomenon where an AI language model confidently generates information that is factually incorrect or entirely fabricated. In voice agent contexts, hallucinations are a critical safety risk — mitigating them requires grounding responses in verified knowledge bases, retrieval-augmented generation, and human-review workflows.
Read definition →
HIPAA
The Health Insurance Portability and Accountability Act — a U.S. federal law that sets national standards for protecting sensitive patient health information (PHI). AI voice agents used in healthcare must follow HIPAA rules around data encryption, access controls, audit logging, and business associate agreements (BAAs) with technology vendors.
Read definition →
Intent Detection
The process by which an AI determines what a caller is trying to accomplish from their spoken or typed input. Accurate intent detection is the foundation of useful voice agents — it routes the conversation to the correct workflow, prevents misunderstandings, and determines when to escalate to a human representative.
Read definition →
IVR (Interactive Voice Response)
A legacy telephony technology that presents callers with pre-recorded audio menus and collects input via keypad presses or basic speech commands. Unlike modern AI voice agents, traditional IVR systems follow rigid scripts, cannot handle natural language, and frequently frustrate callers who fall outside expected input patterns.
Read definition →
Latency
The delay between when a caller finishes speaking and when the AI voice agent begins its response. Low latency (under 800ms) is essential for natural-feeling conversations — high latency creates awkward pauses that signal to callers that they are interacting with a machine rather than a person.
Read definition →
LLM (Large Language Model)
A deep learning model trained on massive text corpora to understand and generate human language. LLMs such as GPT-4, Claude, and Gemini serve as the reasoning core of modern AI voice agents — responsible for comprehending caller intent, formulating accurate responses, and adapting tone to context.
Read definition →
NLP (Natural Language Processing)
A branch of artificial intelligence focused on enabling computers to understand, interpret, and generate human language. NLP is the umbrella discipline that includes intent detection, entity extraction, sentiment analysis, and dialogue management — all foundational to voice agent performance.
Read definition →
NLU (Natural Language Understanding)
The branch of artificial intelligence focused on extracting meaning from human language — identifying intent, entities, and sentiment from speech or text. NLU is the foundation of intent detection in AI voice agents; it allows systems to grasp what a caller wants without relying on rigid menus or keyword matching.
Read definition →
NPS (Net Promoter Score)
A customer loyalty metric calculated from responses to a single question: "How likely are you to recommend this company to a friend or colleague?" on a 0–10 scale. NPS segments respondents into Promoters (9–10), Passives (7–8), and Detractors (0–6), and is widely used in customer experience strategy to benchmark satisfaction and track improvements over time.
Read definition →
Occupancy
A workforce management metric measuring the percentage of an agent's available time spent actively working on customer interactions (calls, emails, chats) versus idle time or administrative tasks. High occupancy rates can improve cost efficiency but risk agent burnout; effective WFM balances occupancy targets with quality and well-being.
Read definition →
Omnichannel
A customer service strategy that seamlessly integrates voice, email, chat, SMS, and social media into a single experience — where customers can start on one channel and continue on another without repeating information. AI voice agents fit into omnichannel ecosystems by handling voice requests and integrating with other channels via shared CRM data and workflows.
Read definition →
PHI (Protected Health Information)
Any information in a medical record or health plan that can be used to identify an individual patient — including name, date of birth, Social Security number, insurance details, and health conditions. HIPAA mandates strict controls on PHI collection, storage, and access. AI voice agents handling healthcare calls must encrypt PHI, audit access logs, and comply with Business Associate Agreement (BAA) obligations.
Read definition →
Power Dialer
An automated dialing technology that dials the next number on an outbound list as soon as an agent becomes available, eliminating idle time between calls. Power dialers present a live caller to the agent within one to two rings, keeping agents continuously engaged and maximizing calls per hour — though requiring careful compliance management to avoid TCPA violations.
Read definition →
Preview Dialer
An outbound dialing system that displays a prospect's profile, previous interaction history, and call notes to an agent before the call is placed. The agent reviews the information and then manually initiates the call, allowing for more personalized and informed conversations. Preview dialers are commonly used in sales and collection environments where context and timing matter.
Read definition →
Predictive Dialer
An automated outbound calling system that uses algorithms to predict agent availability and dials multiple numbers simultaneously, connecting only the live calls to available agents. Predictive dialers maximize call connect rates and agent efficiency but carry regulatory risk — the system must not trigger TCPA violations by creating excessive abandoned calls.
Read definition →
Progressive Dialer
An automated dialing system that dials one number per available agent, waiting for the call to connect before dialing the next number. Progressive dialers are less aggressive than predictive dialers and carry lower TCPA compliance risk because they dial one-to-one with agent availability, resulting in fewer abandoned calls.
Read definition →
RAG (Retrieval-Augmented Generation)
A technique that grounds LLM responses in verified external knowledge by retrieving relevant documents or database records before generating a reply. RAG dramatically reduces hallucinations and keeps voice agents factually accurate by ensuring responses are based on current, authoritative source material rather than model memory alone.
Read definition →
Ringless Voicemail
A technology that delivers a pre-recorded audio message directly to a recipient's voicemail inbox without causing their phone to ring. Ringless voicemail is used for outbound campaigns — appointment reminders, lead nurturing, and re-engagement — and must comply with TCPA regulations governing consent and opt-out mechanisms.
Read definition →
RTP (Real-time Transport Protocol)
A network protocol that enables the real-time transmission of voice and video data across the internet with minimal latency and packet loss. RTP is the foundation of VoIP and cloud-based voice platforms; it ensures that spoken audio in AI voice agent calls and softphone conversations arrives on time and in order.
Read definition →
STT (Speech-to-Text)
The technology that converts spoken audio into written text transcripts. Also called Automatic Speech Recognition (ASR), STT is the first stage of every voice AI pipeline — accuracy here directly determines how well the downstream intent detection and response generation stages perform.
Read definition →
Schedule Adherence
A workforce management metric measuring how closely an agent adheres to their assigned schedule — tracking whether they clock in on time, take breaks only during assigned windows, and avoid unscheduled absences. High schedule adherence ensures consistent coverage, prevents wait time spikes, and is often tied to agent performance reviews and incentive programs.
Read definition →
Shrinkage
A workforce management term referring to any time an agent is being paid but not available to handle customer interactions — including breaks, lunch, training, coaching, meetings, and sick time. Shrinkage is typically expressed as a percentage of scheduled hours and must be factored into staffing models to ensure adequate coverage for customer demand.
Read definition →
SIP (Session Initiation Protocol)
A network protocol used to initiate, maintain, and terminate voice, video, and messaging sessions across internet networks. SIP is the backbone of modern VoIP and cloud telephony systems — including many AI voice agent platforms. It handles signaling (call setup and teardown) separately from media transmission, enabling flexible and scalable voice infrastructure.
Read definition →
Softphone
A software application installed on a computer or mobile device that enables phone calls over the internet (VoIP) without requiring a traditional desk phone. Softphones support voice, video, messaging, and screen sharing, and are essential tools for remote contact center agents, field technicians, and distributed teams working with cloud-based telephony platforms.
Read definition →
STIR/SHAKEN
A framework and set of protocols designed to authenticate the origin of phone calls and prevent spoofing and illegal robocalls. STIR (Secure Telephone Identity Revisited) provides cryptographic verification of a calling number, while SHAKEN (Signature Handling for Authenticated Identity Establishment) defines the implementation standards. Compliance is mandatory for telecom carriers and increasingly expected of outbound calling platforms.
Read definition →
TTS (Text-to-Speech)
The technology that converts written text into synthesized spoken audio. Modern neural TTS engines produce voices that are nearly indistinguishable from human speech, with natural prosody, appropriate pausing, and expressive intonation — critical for creating voice agents that callers find trustworthy and pleasant to interact with.
Read definition →
TCPA (Telephone Consumer Protection Act)
A U.S. federal law restricting unsolicited telemarketing calls, auto-dialed calls, pre-recorded messages, and text messages to consumers. AI voice agent deployments for outbound calling must obtain prior express written consent, honor do-not-call lists, and provide clear opt-out mechanisms to remain TCPA compliant.
Read definition →
Webhook
A method for one application to send real-time event notifications to another via an HTTP request. In AI voice agent workflows, webhooks fire at key moments — call started, intent matched, appointment booked, call ended — enabling immediate downstream actions such as CRM updates, Slack alerts, or follow-up SMS delivery.
Read definition →
Workflow Automation
The use of software to execute a sequence of tasks automatically based on triggers, conditions, and rules — without requiring manual human steps. AI voice agents embed workflow automation to handle post-call actions: updating records, sending confirmation emails, scheduling follow-up calls, and routing leads to the right sales rep.
Read definition →
WFM (Workforce Management)
The discipline of forecasting customer demand, scheduling agents to meet that demand, and monitoring real-time adherence to schedules. WFM teams use metrics like AHT, occupancy, shrinkage, and abandonment rate to optimize staffing levels, budget spend, and customer service quality. Effective WFM balances efficiency with agent well-being.
Read definition →