Appendix A from The AI Contact Center Handbook by Sho Shimoda. Available on Amazon.
How to read this glossary
The contact center industry runs on acronyms; so do AI and compliance. What follows is a single alphabetical A–Z glossary covering all three, sorted strictly by the term itself — an acronym files under its own letters, not its expansion, so AHT is under A, PCI DSS under P, and XOP under X. Where a term is treated in depth in a specific chapter or appendix, the reference points there; where it recurs across the book, the reference points to its first substantive treatment.
In plain English. Skim the letter you need. If a term references a chapter, that is where the concept is unpacked. If it references another appendix, that is where the working tool lives.
A
| Term | Definition |
|---|
| Abandonment Rate | The percentage of inbound calls where the customer hangs up before an agent answers. Rises sharply once expected wait time exceeds roughly 90 seconds. Directly correlated with customer effort and lost revenue. See Chapter 1. |
| ACD (Automatic Call Distributor) | The software or hardware that decides which agent's phone rings when a customer calls in. In a legacy setup, a physical box on your premises; in a modern setup, a service running in the cloud. See Chapter 1. |
| After-Call Work (ACW) | The time an agent spends completing case notes, updating records, and finalizing tasks after a call ends but before they become available for the next one. Increasingly automated by AI note-generation tools. See Chapter 4. |
| Agent Assist | An AI system that runs in real time during a customer conversation, providing suggested responses, knowledge-base articles, and next-best-action recommendations. Sometimes called an in-ear copilot. See Chapter 4. |
| Agent Desktop | The unified interface an agent works from during a customer interaction. In a well-designed operation, a single view of context, history, suggestions, and workflow. In a legacy operation, a euphemism for the eight applications the agent has to alt-tab between. |
| Agentic AI | An AI system that does not just answer questions but takes actions — invoking tools, calling APIs, updating records, and pursuing multi-step goals with minimal human prompting. The step beyond conversational AI. See Chapter 3. |
| AHT (Average Handle Time) | The average duration of a customer interaction from pickup through after-call work. Historically the single most-optimized metric in the industry; increasingly viewed as misleading in an AI-augmented operation where humans handle only the hardest calls. See Chapter 6. |
| API (Application Programming Interface) | The standard way for two pieces of software to talk to each other. When a CCaaS platform "has APIs," it means developers can programmatically read and write data, trigger events, and build custom integrations. See Chapter 2. |
| APPI (Act on the Protection of Personal Information) | Japan's national privacy law, first enacted in 2003 and substantially amended in 2020 and 2022. Comparable in scope to GDPR though less prescriptive on some points. See Appendix C. |
| ASV (Approved Scanning Vendor) | Under PCI DSS, a company approved by the PCI Security Standards Council to perform the external vulnerability scans required for merchants and service providers. See Appendix C. |
| Attribution | The practice of assigning credit for a business outcome to specific interactions along the customer journey. Central to measuring the value of a CX investment, and notoriously hard to do rigorously. |
B
| Term | Definition |
|---|
| Bake-Off | Industry slang for a vendor selection process in which multiple vendors give competing demos to the same buyer, often in a single day. Rarely produces the differentiation buyers expect. See Chapter 12. |
| Blended Agent | An agent who handles interactions across multiple channels — voice, chat, email, messaging — rather than specializing in one. |
| Blended Learning | A training approach combining self-paced digital coursework with instructor-led sessions and hands-on practice. Increasingly standard for multi-month reskilling programs. |
| BPO (Business Process Outsourcing) | The practice of contracting out a business function to a third-party provider, often located in a lower-cost geography. The Philippines, India, and Latin America are major hubs. See Chapter 11. |
| Business Continuity Plan (BCP) | Documented procedures for maintaining critical operations during and after a disruptive event. For contact centers, addresses staffing failover, systems failover, and customer communication during outages. |
| BYOK (Bring Your Own Key) | An arrangement in which the customer supplies and controls the cryptographic keys used to encrypt their data at rest inside a vendor's cloud. Common in regulated industries. See Chapter 9. |
C
| Term | Definition |
|---|
| Callback | An option offered to a customer in a queue: rather than hold, provide a number and have the system call back when an agent is available. Modern implementations preserve queue position. See Chapter 6. |
| CAMARA | An open-source telecommunications API project run by the Linux Foundation, exposing network-level intelligence (device location, SIM swap detection, number verification) that XOP platforms use for real-time fraud and identity decisions. See Chapter 5. |
| CCaaS (Contact Center as a Service) | The cloud-hosted version of the contact center. Major vendors include Genesys, Five9, NICE, Amazon Connect, Talkdesk, and Content Guru. See Chapter 2. |
| CDP (Customer Data Platform) | Software that unifies customer data from multiple sources into a single, queryable profile. Increasingly being displaced by XOP platforms that add real-time decisioning. See Chapter 5. |
| CES (Customer Effort Score) | A metric asking the customer how much effort it took to resolve their issue. Correlates more strongly with loyalty behavior than CSAT for support interactions. See Chapter 1. |
| Champion / Challenger | A testing methodology in which a current production configuration (the champion) is compared to alternatives (challengers) using live traffic. Standard for testing AI prompts, routing rules, and conversation flows. |
| CLV (Customer Lifetime Value) | The total revenue a customer is expected to generate over the duration of their relationship. Increasingly used in real-time routing — high-CLV customers get shorter queues and more senior agents. See Chapter 7. |
| Containment | The percentage of interactions an automated agent handles end-to-end without escalating. The primary business metric for a deployed AI agent. See Chapter 3. |
| Cost per Contact | The fully loaded operational cost divided by the number of customer contacts. The primary efficiency metric in most contact center P&Ls. |
| CPaaS (Communications Platform as a Service) | Developer-facing building blocks — SMS, voice, video, and chat APIs. Twilio and Vonage lead. See Chapter 2. |
| CRM (Customer Relationship Management) | Software that stores customer records and interaction history. Salesforce, Microsoft Dynamics, HubSpot, and Zendesk are common. See Chapter 1. |
| CSAT (Customer Satisfaction Score) | A metric asking customers how satisfied they were with a specific interaction. The oldest and most widely used CX metric. See Chapter 6. |
| CVV (Card Verification Value) | The three- or four-digit code on a payment card used to verify physical possession. Under PCI DSS, CVV must never be stored after authorization. See Chapter 9. |
D
| Term | Definition |
|---|
| Deepfake | A synthetic audio or video artifact generated by AI to impersonate a real person. Primary contact center risk: voice deepfakes defeating voice biometrics. See Chapter 10. |
| DPIA (Data Protection Impact Assessment) | A structured risk assessment required under GDPR (and increasingly other regimes) for high-risk processing. Voice AI usually triggers one. See Appendix B. |
| DTMF (Dual-Tone Multi-Frequency) | The technical name for the beeps your phone makes when you press keys. In a contact center, how customers enter account and card numbers. DTMF masking prevents tones from being recorded during payment capture. See Chapter 9. |
E
| Term | Definition |
|---|
| Erlang C | A queueing-theory formula, developed by Agner Krarup Erlang in 1917, used to calculate the number of agents required to handle a given call volume at a target service level. See Chapter 6 and Appendix F. |
| Erlang O | A more recent generalization for asynchronous, concurrent, and blended channels where a single agent handles multiple interactions simultaneously. See Appendix F. |
| Escalation | The transfer of an interaction from one agent (human or AI) to another. See Chapter 4. |
| EU AI Act | The European Union's 2024 regulation on artificial intelligence. Contact center workforce-management AI and biometric identification fall into the "high-risk" category. See Chapter 8. |
F
| Term | Definition |
|---|
| FCR (First Contact Resolution) | The percentage of interactions fully resolved on the first contact, without a callback or transfer. Widely considered the single best predictor of both customer satisfaction and operational cost. See Chapter 1. |
| Federation | An identity architecture in which authentication is delegated to a trusted external provider (SAML, OAuth, OIDC). What enables SSO across an enterprise stack. |
| First Party Fraud | Fraud committed by the account holder against the company — for example, disputing a legitimate transaction to receive a refund. |
G
| Term | Definition |
|---|
| GDPR (General Data Protection Regulation) | The EU's comprehensive privacy law, in force since 2018. Penalties can reach 4% of global annual turnover. See Chapter 8. |
| Guardian Agent | An AI system whose job is to watch other AI systems in real time and intervene when they are about to take an unauthorized, unsafe, or non-compliant action. See Chapter 10. |
H
| Term | Definition |
|---|
| HIPAA (Health Insurance Portability and Accountability Act) | The U.S. federal law governing the privacy and security of protected health information. Applies to healthcare providers, insurers, and their business associates. See Chapter 8. |
I
| Term | Definition |
|---|
| ICMI (International Customer Management Institute) | A long-established U.S. training, research, and community organization for contact center professionals. |
| Intent Recognition | The AI capability of identifying what a customer actually wants from a natural-language utterance. See Chapter 3. |
| IVA (Intelligent Virtual Agent) | An AI-powered virtual agent capable of holding a natural-language conversation and completing transactions end-to-end. The modern successor to the traditional IVR menu. See Chapter 3. |
| IVR (Interactive Voice Response) | The automated menu system that answers your call and asks you to press 1 for billing. Historically the most-hated part of the customer experience. See Chapter 1. |
J
| Term | Definition |
|---|
| Journey Analytics | The practice of measuring customer behavior across the full sequence of interactions with a company, rather than treating each interaction as a standalone event. See Chapter 7. |
K
| Term | Definition |
|---|
| Knowledge Base | The repository of articles, policies, and reference materials that agents (human and AI) consult. Modern implementations use RAG to retrieve relevant articles automatically. See Chapter 3. |
| KPI (Key Performance Indicator) | A metric selected as a headline measure of operational performance. Common ones: AHT, CSAT, FCR, service level, abandonment rate. See Chapter 6. |
L
| Term | Definition |
|---|
| LGPD (Lei Geral de Proteção de Dados) | Brazil's general data protection law, in force since 2020. Structurally similar to GDPR. See Appendix C. |
| LLM (Large Language Model) | A machine learning model trained on very large text corpora, capable of understanding and generating human-like language. GPT-4, Claude, Gemini, and Llama are examples. See Chapter 3. |
M
| Term | Definition |
|---|
| MCP (Model Context Protocol) | An open protocol, introduced by Anthropic in 2024, that standardizes how AI models connect to external tools and data. MCP is to AI what USB is to peripherals. See Appendix D. |
| MFA (Multi-Factor Authentication) | A security control requiring two or more independent credentials before access is granted. Baseline requirement in most vendor security reviews. See Chapter 9. |
| Model Drift | The gradual degradation of an AI model's performance over time as real-world data shifts away from training data. In a contact center, drift shows up as declining containment rates and rising escalations. |
| Multimodal | An AI system capable of processing more than one type of input — text, voice, images, video — in a single interaction. |
N
| Term | Definition |
|---|
| NPS (Net Promoter Score) | A loyalty metric based on a single question about likelihood to recommend. Widely used but frequently criticized for oversimplification. See Chapter 7. |
O
| Term | Definition |
|---|
| Observability | The practice of instrumenting a system so its internal state can be inferred from outputs — logs, metrics, traces. For agentic AI, includes model-level metrics: token usage, retrieval quality, safety-flag frequency. |
| Occupancy | The percentage of an agent's scheduled time actually spent handling interactions. Industry targets are 85–90%; above that, burnout rises sharply. See Chapter 6. |
| OIDC (OpenID Connect) | An identity layer built on top of OAuth 2.0, used to verify identity and obtain basic profile information from an identity provider. The modern web equivalent of SAML. |
| Omnichannel | The capability of interacting with customers seamlessly across multiple channels with continuity of context. See Chapter 7. |
| Optichannel | A refinement of omnichannel: routing each interaction to the channel that will resolve it best, given the context. See Chapter 7. |
P
| Term | Definition |
|---|
| P&L (Profit and Loss) | The financial statement summarizing revenues, costs, and expenses over a defined period. The frame in which any AI business case has to be defended. |
| PBX (Private Branch Exchange) | The private telephone system that routes calls within an organization. See Chapter 1. |
| PCI DSS (Payment Card Industry Data Security Standard) | The security standard for any organization that handles credit card data. Currently at version 4.0. See Chapter 9 and Appendix C. |
| PII (Personally Identifiable Information) | Any information that can be used to identify a specific individual. Most privacy regulations are structured around the protection of PII. |
| Post-Call Summary | An AI-generated summary of a completed interaction, typically written into the CRM case record. Replaces the manual after-call work agents used to do. |
| Prompt Injection | An attack in which a malicious user embeds instructions in an input to an AI system with the intent of overriding the system's original instructions. The AI equivalent of SQL injection. See Chapter 10. |
Q
| Term | Definition |
|---|
| QA (Quality Assurance) | The function responsible for evaluating the quality of agent interactions. Historically based on sampling roughly 2% of calls; increasingly automated by AI platforms monitoring 100% of interactions. See Chapter 6. |
R
| Term | Definition |
|---|
| RAG (Retrieval-Augmented Generation) | A pattern in which an AI model retrieves relevant documents from a knowledge base before generating its response, grounding the answer in specific source material. The dominant architecture for enterprise Q&A in 2026. See Chapter 3. |
| RFP (Request for Proposal) | A formal document a buyer issues to prospective vendors describing requirements and evaluation criteria. See Chapter 12. |
| Right-to-Be-Forgotten | The customer's legal right, under GDPR and analogous frameworks, to require deletion of their personal data. Cascades into recordings, transcripts, AI training datasets, and derived analytics. |
| RPA (Robotic Process Automation) | Software that automates repetitive digital work by scripting mouse clicks and keystrokes. Older than agentic AI and largely being absorbed by it. See Chapter 11. |
S
| Term | Definition |
|---|
| SAML (Security Assertion Markup Language) | An XML-based open standard for exchanging authentication and authorization data. One of the two main protocols (alongside OIDC) behind enterprise SSO. |
| SAQ (Self-Assessment Questionnaire) | Under PCI DSS, the questionnaire that merchants at Levels 2–4 use to attest to their compliance. See Appendix C. |
| Sentiment Analysis | AI-driven classification of a customer's emotional state during an interaction, usually as positive, neutral, or negative. |
| Service Level (SL) | The percentage of calls answered within a defined target time, typically expressed as X/Y — for example, "80/20" means 80% of calls answered within 20 seconds. See Chapter 6. |
| Shrinkage | The gap between scheduled agent hours and actual productive hours, once you subtract breaks, meetings, training. Typically 25–35%. See Chapter 6. |
| SIEM (Security Information and Event Management) | A platform that aggregates security-relevant logs and events from across an enterprise and applies correlation rules to detect threats. |
| SIP (Session Initiation Protocol) | The internet protocol used to set up voice, video, and messaging sessions. The plumbing under most modern VoIP-based contact centers. See Chapter 2. |
| Skills-Based Routing | The practice of routing an interaction to the agent best equipped to handle it, based on skills and interaction requirements. Standard in modern CCaaS. |
| SLA (Service Level Agreement) | A contract term specifying the level of service a vendor commits to deliver, and the penalties if they fail. See Chapter 12. |
| Speech Analytics | AI-driven analysis of the content of voice calls — transcription, keyword detection, sentiment scoring, compliance monitoring. See Chapter 6. |
| SSO (Single Sign-On) | A system that lets users log into multiple applications using a single set of credentials. See Chapter 1. |
| STIR/SHAKEN | A set of protocols, mandated by the FCC and required by carriers in a number of jurisdictions, for authenticating the caller ID of voice calls. See Chapter 10. |
| System Prompt | The persistent instructions given to an AI model that shape its behavior across all conversations. Writing a good system prompt is a specialized skill. |
T
| Term | Definition |
|---|
| TCO (Total Cost of Ownership) | The full cost of a system over its useful life, including licenses, hardware, integration, training, ongoing operations, and eventual replacement. See Appendix E. |
| TCPA (Telephone Consumer Protection Act) | The U.S. federal law restricting telemarketing, auto-dialed calls, prerecorded messages, and unsolicited texts. Statutory damages of $500–$1,500 per violation. See Chapter 8. |
U
| Term | Definition |
|---|
| UCaaS (Unified Communications as a Service) | The cloud-hosted version of the internal collaboration stack — voice, video, messaging, screen sharing for employees. Microsoft Teams, Zoom, and RingCentral lead. See Chapter 2. |
V
| Term | Definition |
|---|
| Voice Biometrics | Authentication of a caller based on the acoustic characteristics of their voice, typically as a mathematical voiceprint. The frontline defense against deepfake fraud in 2026. See Chapter 10. |
| VoIP (Voice over Internet Protocol) | Voice communication carried over an IP data network rather than the traditional circuit-switched telephone network. The technology that made cloud-based contact centers possible. See Chapter 2. |
W
| Term | Definition |
|---|
| Wallboard | A large display, historically mounted on the wall of the contact center floor, showing real-time operational metrics. |
| Warm Transfer | A transfer in which the sending agent briefly speaks with the receiving agent before handing off the customer, providing context. Preferable to a cold transfer. See Chapter 1. |
| WEM (Workforce Engagement Management) | The evolved form of WFM. Includes workforce management, quality management, performance management, coaching, and often gamification. See Chapter 6. |
| WFM (Workforce Management) | The narrower predecessor to WEM: scheduling agents, forecasting call volumes, and tracking adherence. See Chapter 6. |
X
| Term | Definition |
|---|
| XOP (Experience Orchestration Platform) | A platform that unifies real-time customer data, decisioning logic, and network-level intelligence to make in-the-moment routing and treatment decisions. The next-generation replacement for the traditional CDP. See Chapter 5. |
Z
| Term | Definition |
|---|
| Zero-Party Data | Data that a customer knowingly and voluntarily provides to a company, as distinct from first-party (observed) and third-party (obtained from external sources). Increasingly valued as third-party data becomes harder to acquire. |
One-line takeaway. Contact center and AI vocabulary is a moving target — treat this glossary as a working reference, and check the chapter link for anything you plan to build a decision around.
How to use this appendix
Reach for the glossary when a term arrives in a meeting, a vendor deck, or a compliance document that you have not seen before. The one-sentence definition is enough to keep the conversation moving; the chapter or appendix reference is where you go afterward to understand the concept in enough depth to defend a decision.