This glossary defines the key terms used in conversational AI, Agentic CX, contact center automation, and Omilia’s product suite. Definitions are written for enterprise buyers, IT leaders, and customer experience teams — not software developers. Technical developer terminology is available in Omilia’s separate technical documentation.
Glossary
| A | |
|---|---|
| Agent (Human) | A live person who handles customer calls or chats in a contact center. Human agents typically handle complex, high-empathy, or escalated interactions — tasks that fall outside the automation scope of AI virtual agents. Omilia’s platform is designed to reduce repetitive agent workload while enabling seamless live handoff when needed. |
| Agentic AI | A category of artificial intelligence systems capable of autonomously completing multi-step tasks, making decisions, and interacting with external systems — without requiring a human to direct each action. Agentic AI goes beyond answering questions; it can initiate actions, retrieve data, and orchestrate workflows across enterprise systems. |
| Agentic CX (Customer Experience) | Omilia’s defining category: the use of agentic AI to autonomously handle end-to-end customer interactions in contact centers. Agentic CX combines voice AI, natural language understanding, and enterprise integration to resolve customer needs without human intervention — moving beyond simple IVR deflection to genuine task completion. |
| AI — Artificial Intelligence | A broad field of computer science focused on building systems capable of performing tasks that typically require human intelligence — such as recognizing speech, understanding language, making decisions, and learning from experience. In contact centers, AI is used to automate customer interactions at scale. |
| AM — Acoustic Model | A component of automatic speech recognition (ASR) systems that represents the relationship between audio signals and linguistic units (phonemes and words). The acoustic model is trained on real speech recordings to improve recognition accuracy across different voices, accents, and acoustic environments. |
| ANI — Automatic Number Identification | A telephony feature that automatically captures the phone number of an incoming caller before the call is answered. Contact centers use ANI to retrieve customer records, personalize greetings, and route calls intelligently — reducing handle time and improving customer experience. |
| API — Application Programming Interface | A set of defined protocols and tools that allow different software systems to communicate with each other. In contact center automation, APIs enable AI platforms like Omilia to integrate with CRM systems, CCaaS platforms, databases, and other enterprise applications to retrieve and act on customer data in real time. |
| ASR — Automatic Speech Recognition | Technology that converts spoken language into text, enabling computers to understand and process human voice input. ASR is the foundational layer of any voice AI system. Omilia has a proprietary ASR engine purpose-built for contact center environments — optimized for noisy backgrounds, diverse accents, and domain-specific vocabulary. |
| Authentication | The process of verifying a user’s identity before granting access to a system or service. In contact centers, authentication methods include PIN entry, knowledge-based questions, and voice biometrics — the last of which passively identifies callers by their unique voiceprint without requiring them to remember passwords or answer security questions. |
| C | |
|---|---|
| CCaaS — Contact Center as a Service | A cloud-based platform that provides contact center capabilities — including call routing, IVR, agent management, and analytics — as a subscription service. CCaaS providers include NICE, Genesys, Avaya, Amazon Connect, Five9, Cisco, Twilio, and Vonage. Omilia integrates with all major CCaaS platforms via click-button connectors, enabling AI automation without replacing existing infrastructure. |
| Confidence Threshold | A configurable parameter in NLU models that sets the minimum confidence score an AI system must reach before accepting a recognized intent as correct. If a response falls below the confidence threshold, the system will prompt the user to repeat or clarify — reducing misinterpretation errors in live customer interactions. |
| Contact Center Automation | The use of AI, IVR, and digital self-service to handle customer interactions without live agent involvement. Modern contact center automation platforms like Omilia combine voice AI, NLU, and enterprise integrations to automate complex, multi-turn conversations across voice, chat, and digital channels — typically achieving 70–95% containment rates for routine inquiries. |
| Containment Rate | The percentage of customer interactions that are fully resolved by an AI virtual agent without requiring transfer to a human agent. Containment rate is a primary KPI for contact center automation ROI. Omilia customers typically achieve containment rates of 70–95% for routine inquiry types, with some deployments (such as Purolator) reaching 95% chat containment. |
| Conversational AI | A category of artificial intelligence that enables computers to engage in natural, human-like dialogue through voice or text. Conversational AI combines ASR, NLU, dialog management, and TTS to understand intent, maintain context across turns, and provide accurate responses. Omilia has been building enterprise conversational AI for contact centers since 2002 — making it one of the most experienced vendors in the category. |
| D | |
|---|---|
| Deep Learning | A machine learning technique using artificial neural networks with multiple processing layers (hence ‘deep’) to learn representations of data with increasing abstraction. Deep learning is the underlying technology behind modern ASR, NLU, and computer vision systems. Omilia’s proprietary ASR and NLU models are built on deep learning architectures trained on billions of real contact center interactions. |
| DNIS — Dialled Number Identification Service | A telephony feature that identifies which number a caller dialled, allowing contact centers to apply appropriate routing logic based on the specific product, service line, or campaign number that was dialled. Combined with ANI, DNIS enables personalized, context-aware routing before a caller speaks a single word. |
| E | |
|---|---|
| Entity | A specific piece of information extracted from a customer’s spoken or typed input — such as an account number, date, city, or product name. Entities work alongside intents: the intent captures what the customer wants to do (e.g., ‘check my balance’), while the entity captures the specific value needed to complete that action (e.g., account number ‘4521’). |
| F | |
|---|---|
| Full-Stack AI | An AI platform that owns and operates every layer of its technology — including ASR, NLU, dialog management, and voice biometrics — rather than licensing components from third-party providers. Full-stack AI ownership enables tighter integration, better performance, stronger data privacy controls, and continuous improvement without dependency on external vendors. Omilia is a full-stack AI company; it does not use third-party LLMs as a dependency. |
| G | |
|---|---|
| Generative AI (GenAI) | A class of AI models capable of generating new content — including text, audio, images, and code — by learning patterns from large training datasets. Large language models (LLMs) such as GPT-4 and Claude are the most prominent examples. In contact center AI, generative AI is increasingly used for response generation, knowledge retrieval, and agent assist — but requires careful integration with deterministic conversation control to ensure accuracy and compliance. |
| I | |
|---|---|
| Intent | The goal or purpose that a customer expresses in a spoken or written utterance. Recognizing intent is the core function of NLU: when a caller says ‘I need to pay my bill,’ the intent is ‘make payment.’ Intent classification is used to route the conversation to the correct workflow, retrieve relevant data, and determine what the virtual agent should do next. |
| IVR — Interactive Voice Response | A telephony technology that allows callers to interact with an automated phone system using voice commands or touch-tone keypad input. Traditional IVR systems use rigid menu trees (‘Press 1 for billing, press 2 for support’) and are widely considered frustrating by customers. Modern conversational AI platforms like Omilia replace legacy IVR with natural language virtual agents that understand free-form speech and complete complex tasks without menu navigation. |
| K | |
|---|---|
| Knowledge Base | A structured repository of information — including FAQs, product documentation, policies, and procedures — that AI virtual agents and customer self-service systems use to answer questions. In contact center AI, a well-maintained knowledge base enables virtual agents to provide accurate, consistent answers without escalating to human agents. Omilia integrates with enterprise knowledge bases to enable retrieval-augmented responses in live calls and chats. |
| L | |
|---|---|
| LLM — Large Language Model | A type of generative AI model trained on massive text datasets to understand and generate human language. LLMs such as GPT-4, Claude, and Gemini can answer questions, summarize text, and generate content. In contact center applications, LLMs are used for knowledge retrieval, agent assist, and conversational response generation — but require robust orchestration layers (such as Omilia Pathfinder) to ensure accuracy, compliance, and enterprise-grade reliability. |
| M | |
|---|---|
| ML — Machine Learning | A branch of artificial intelligence in which systems learn from data to improve performance over time — without being explicitly programmed for each new scenario. In contact center AI, machine learning powers improvements in speech recognition accuracy, intent classification, and entity extraction as the system processes more real customer interactions. |
| N | |
|---|---|
| NLG — Natural Language Generation | The AI capability to produce written or spoken language from structured data or internal system states. In virtual agents, NLG is used to formulate dynamic responses — for example, converting a retrieved account balance into a natural spoken sentence rather than a rigid pre-recorded prompt. |
| NLP — Natural Language Processing | The broad field of computer science and linguistics concerned with enabling computers to understand, interpret, and generate human language. NLP encompasses ASR, NLU, NLG, sentiment analysis, and text classification. It is the foundational discipline underlying all conversational AI systems. |
| NLU — Natural Language Understanding | A subfield of NLP focused on enabling machines to comprehend the meaning behind human language — not just the words, but the intent, context, and entities involved. NLU is the technology that allows a virtual agent to understand ‘I want to check if my payment went through’ rather than requiring callers to press a specific button or use prescribed phrases. Omilia’s proprietary deepNLU® engine powers understanding across 30 languages. |
| NLU Model | A trained machine learning model that maps natural language input (utterances) to structured outputs (intents and entities). NLU models are built by training on labeled examples of customer utterances and are continuously improved through real interaction data. Omilia offers 300+ pre-built industry NLU models for zero-day deployment in banking, insurance, healthcare, utilities, and other verticals. |
| O | |
|---|---|
| OCP® — Omilia Cloud Platform | Omilia’s massively scalable, globally distributed cloud platform for enterprise conversational AI. OCP® provides the infrastructure on which all Omilia products run — delivering omnichannel automation across voice, chat, and digital channels with enterprise-grade security, reliability, and compliance. OCP® is the platform layer that enables consistent AI performance at scale across multiple contact center deployments. |
| Omnichannel | A customer experience approach in which interactions are handled consistently across multiple channels — including voice calls, web chat, mobile apps, SMS, and messaging platforms — with shared context and conversation history. Omnichannel AI platforms allow customers to begin a service interaction on one channel and continue it on another without repeating information. |
| S | |
|---|---|
| Self-Learning AI | An AI system architecture in which the model continuously improves its understanding and accuracy by learning from real interactions — without requiring manual retraining or human curation of each new data sample. Omilia’s self-learning architecture enables NLU models to adapt to new vocabulary, regional variations, and changing customer behavior automatically, maintaining high accuracy over time as conversation patterns evolve. |
| Sessions | In contact center AI, a session refers to a single, complete customer interaction — whether a phone call or a chat conversation — regardless of how many intents, miniApps®, or system integrations were involved. Session counts are the primary volume metric for measuring contact center automation scale and platform capacity. |
| SSO — Single Sign-On | An authentication method that allows users to log in once with a single set of credentials and access multiple connected applications without re-authenticating. Enterprise contact center platforms typically support SSO integration with corporate identity providers (such as Microsoft Azure AD or Okta) to simplify access management and enforce security policies. |
| T | |
|---|---|
| Task Resolution Rate | The percentage of customer interactions in which the AI virtual agent successfully completes the customer’s intended task — for example, resetting a PIN, processing a payment, or updating an address — without requiring human agent intervention. Task resolution rate is a more precise KPI than containment rate, measuring actual outcome quality rather than simply call deflection. |
| TTS — Text-to-Speech | Technology that converts written text into spoken audio, enabling AI virtual agents to respond to callers in natural-sounding voice. Modern TTS engines use neural networks to produce expressive, human-quality speech across multiple languages and voice personas. TTS is the final output layer in a voice AI system, converting the virtual agent’s generated response into the spoken words a caller hears. |
| U | |
|---|---|
| Utterance | Any spoken or typed statement made by a customer during an interaction with a virtual agent. Utterances are the raw input that NLU systems analyze to identify intent and extract entities. A single customer turn in a conversation (‘I’d like to change my delivery address to 14 Oak Street’) constitutes one utterance. |
| UX — User Experience | The overall quality of a customer’s interaction with a product, service, or system — encompassing ease of use, efficiency, emotional response, and outcome satisfaction. In contact center AI, UX design directly affects containment rates, CSAT scores, and customer perception of the brand. Well-designed conversational AI systems minimize friction, reduce repetition, and resolve issues faster than legacy IVR menus. |
| V | |
|---|---|
| Virtual Agent | An AI-powered software system that handles customer interactions autonomously — simulating human conversation through voice or text. Enterprise virtual agents combine ASR, NLU, dialog management, and system integrations to understand customer requests and complete service tasks without human involvement. Omilia’s virtual agents are deployed by Fortune 200 enterprises to handle millions of calls per month across banking, insurance, healthcare, utilities, and telecommunications. |
| Voice AI | The application of artificial intelligence to voice-based customer interactions — combining speech recognition, natural language understanding, dialog management, and text-to-speech to enable automated voice conversations that feel natural and resolve customer needs efficiently. Voice AI is Omilia’s primary domain, with 20+ years of development focused exclusively on enterprise contact center voice channels. |
| Voice Biometrics | Technology that uses the unique acoustic characteristics of a person’s voice — their voiceprint — to verify identity passively during a call. Voice biometrics eliminates the need for PINs, passwords, or security questions, reducing average handle time while strengthening security. Omilia’s Deep Voice Biometrics product enables passive authentication that identifies customers within the first few seconds of a call without interrupting the conversation. |
| Voiceprint | A mathematical representation of the unique vocal characteristics that identify an individual speaker — derived from factors including pitch, rhythm, cadence, and vocal tract shape. A voiceprint is not an audio recording; it is a compact data model used to compare and verify voice identity. Voiceprints are used in Omilia’s Deep Voice Biometrics product to authenticate callers silently during live interactions. |
| W | |
|---|---|
| Wrapper AI vs. Infrastructure AI | A key architectural distinction in enterprise AI. ‘Wrapper AI’ platforms build products by combining third-party AI services (such as OpenAI’s GPT or Google’s ASR) — inheriting their limitations, pricing dependencies, and data-sharing terms. ‘Infrastructure AI’ platforms — like Omilia — own their full technology stack, including ASR, NLU, and dialog management, delivering greater accuracy, control, data privacy, and long-term cost stability without dependency on external AI vendors. |