Customer Service and AI Customer Service Glossary

The proportion of customers who leave a service queue before receiving help or support from a business. In telephone support, this generally means callers who disconnect before an agent answers. In chat or messaging, it may refer to customers who exit before the conversation is accepted or completed.

A rising abandonment rate can point to long waiting times, poor queue messaging or customers deciding that the effort required is no longer worthwhile.

The administrative work an agent completes once a customer conversation has ended. This may involve adding case notes, categorising the contact, updating an account or arranging a follow-up, or escalating the customer's query for investigation.

After-call work is commonly shortened to ACW and may also be called wrap-up time. An agent's after-call work (ACW) is commonly measured by the time taken to complete it.

A person responsible for handling customer enquiries, requests, complaints or support cases.

The word is now also used for automated systems, so organisations may specify whether they are referring to a human agent, virtual agent or AI agent.

Software, most likely AI software, that helps an employee while they are speaking or writing to a customer. It can bring up relevant knowledge, suggest wording, highlight compliance requirements or create notes from the conversation.

The employee continues to manage the interaction and decides whether the suggested information should be used.

The period during which an employee is signed into the service platform and ready to accept customer contacts.

Availability is not always the same as paid working time because training, meetings, breaks and administration may temporarily remove an agent from the queue, affecting their availability.

The working experience of people who deliver customer service. It is shaped by workload, systems, management, access to accurate information, training and the level of discretion employees have when helping customers. Low control and high demand in a job role can negatively affect the agent experience.

Agent experience is frequently abbreviated to AX.

The share of an agent's available time spent actively managing customer contacts or completing work directly connected to them.

High occupancy can indicate that staffing is being used effectively. If it remains too high for long periods, employees may have little recovery time between difficult conversations and may experience frontline fatigue.

A form of AI designed to autonomously carry out a goal through several connected steps. Instead of answering one question and stopping, or responding to prompts, an agentic system may gather information, choose an action, interact with another platform and confirm that the task has been completed.

In customer service, this could include checking an order, changing a delivery instruction and sending the customer an updated confirmation.

A structured way for separate software products to communicate with one another.

An application programming interface, usually called an API, might allow a chatbot to retrieve an order status from an ecommerce platform or allow a support system to create a refund in a payment platform.

A proposed form of artificial intelligence that could perform intellectual work across many unrelated areas rather than operating within a limited task. Artificial general intelligence, or AGI, remains a theoretical concept. Current customer service AI is designed for narrower purposes.

A broad category of technology used to perform tasks that involve interpretation, prediction, language, pattern recognition or decision support. Customer service applications include conversation analysis, automated routing, voice recognition, response generation and forecasting.

The use of software, now usually artificial intelligence powered software, to review customer interactions against quality criteria. AQMs usually evaluate compliance, the quality of interactions and may score agent performance. Traditional quality programmes may evaluate a small sample of conversations manually. Automated quality management can examine a much larger set, although human review remains important for context, judgement and disputed results.

Technology that receives incoming calls and assigns them to agents or queues according to routing rules. An automatic call distribution system, usually shortened to ACD, may consider language, department, agent skill, customer status or current queue conditions.

Technology that turns spoken language into written text. Automatic speech recognition, or ASR, is used in call transcription, voice assistants, voice-bots and speech analytics.

The proportion of customer contacts, tasks or processes completed through automated systems. Organisations should state exactly what the figure includes. A contact that stays within an automated system has not necessarily been resolved successfully.

The average amount of operational time used to manage a customer interaction. For telephone service, it is often calculated as:

Total talk time + total hold time + total after-call work ÷ number of handled calls

Average handle time is usually shortened to AHT. It is useful for planning capacity, although treating a low AHT as the main objective can encourage rushed service. A high AHT can sometimes indicate different issues, such as an agent dealing with concurrent complex calls, or an agent who requires further training. Reducing AHT allows each agent to handle more customer interactions within a given period.

The typical length of time between a customer first raising an issue and the issue being completed. This measure may include periods when the case is waiting for another department, a supplier or further information from the customer.

The average time customers wait before an agent answers their call or accepts their live conversation. Average speed of answer is commonly abbreviated to ASA.

The total amount of unresolved work waiting for attention. A backlog may consist of unanswered emails, open support tickets, complaint investigations or cases requiring follow-up. A high backlog over a long period of time may require root cause analysis to identify roadblocks impeding customer resolution rates.

A call supervision function that allows a manager or supervisor to join a live customer interaction, either silently to monitor the call, privately to whisper guidance to the agent, or audibly to speak with both the agent and the customer.

Comparing service performance with a reference point. The comparison point may be against previous results, another team, an industry figure or a competitor. Benchmarks are most useful when the underlying measures have been defined in the same way.

A software program, now usually AI-powered, that carries out an automated activity. Within customer service, the term commonly refers to chatbots, voice-bots or other automated systems that communicate with customers.

The proportion of users who leave an automated conversation before completing the intended journey. This may indicate that the bot was difficult to use, failed to understand the request or required too many steps, or the customer may prefer to interact with the company through an alternative communication method.

The percentage of automated conversations that are fully resolved by a chatbot or an AI agent, rather than being transferred to or escalated to an employee.

The use of advanced technology to carry out recurring operational processes with limited manual handling. Business process automation, often shortened to BPA, may cover several systems and departments. Examples include registering complaints, updating customer records or processing standard refund requests.

A service operation built mainly around telephone conversations. Some call centres also manage other channels, although the term usually places greater emphasis on calls than the broader term contact centre.

The sequence a telephone interaction follows from the moment the customer connects. It may include greetings, identity checks, menu options, routing rules, queues and transfers.

The review of live or recorded customer calls to assess service quality, identify risks and support agent development. Monitoring may be completed by supervisors, quality teams or automated systems.

The capture and storage of a telephone conversation for purposes such as quality review, training, dispute investigation or compliance. Recording must follow applicable privacy, consent and retention requirements.

An option that allows a customer to request a return call rather than continue waiting in a telephone queue. Callbacks can reduce visible queue pressure while allowing customers to use their time elsewhere.

A record created to manage a customer issue that may continue beyond a single conversation with the agent who first interacts with the customer. Cases are often used for complex complaints, investigations, insurance claims, technical faults or requests involving collaboration between several departments.

Case management tracks a customer issue from the initial contact through to final resolution, keeping all related interactions, updates, documents and actions together so agents can easily understand the full history of the case.

A route through which a customer communicates with an organisation. Examples include telephone, email, live chat, online messaging, social media, video support and face-to-face service.

A customer's movement from one service channel to another. A conversation might begin in self-service, move to chat and later continue by telephone. Repeated switching can indicate that earlier channels did not provide enough help to the customer.

The number of customer conversations an agent manages at the same time. Higher concurrency can increase capacity, although excessive levels may reduce attention and response quality.

A text-based system that communicates with users through a conversational interface. Some chatbots follow fixed menus and rules. Others use natural language processing or generative AI to interpret a wider range of questions.

Sometimes called churn rate. Churn is the loss of customers over a defined period. A customer may be counted as churned after cancelling a subscription, closing an account or no longer purchasing from the organisation.

The assignment of information to a category. In customer service, classification may be used to label contact reasons, identify complaints, detect urgency or route work to the appropriate team.

Closed-loop feedback turns customer responses and feedback, from sources such as surveys, into action by identifying service failures, assigning them for follow-up and tracking the outcome. The information gathered is also used to strengthen training and improve how the contact centre operates.

Technology that allows an employee to view or guide a customer's online session with permission. It can be useful when helping someone complete a form, locate a setting or understand and complete a complicated digital process.

The direct transfer of a customer to another employee, department or queue without an introduction or discussion with the receiving agent. It may also be called a blind transfer.

An expression of dissatisfaction in which the customer expects the organisation to respond, investigate or correct something. A complaint may concern an employee, decision, product, process, delay or service failure or another other issue.

The process used to receive, record, investigate and respond to customer complaints. A mature complaint-handling process also looks for recurring causes, or root causes, rather than treating every case as an isolated event.

The volume of complaints compared with a wider business measure, such as total customer interactions. It may be reported per thousand customers, per order, per transaction or as a percentage of all service contacts. Complaint rate can serve as a diagnostic measure of overall satisfaction with product or service quality, process friction or satisfaction with agent service and training.

Following applicable laws, regulations, contractual obligations and organisational policies. Customer service compliance may involve privacy, identity checks, recordkeeping, accessibility, financial disclosure, regulatory adherence, mandatory disclosures, complaint deadlines or other requirements.

This score indicates how certain an AI system is that it has correctly understood a customer's request and selected an appropriate response. When confidence falls below a set threshold, the interaction may be escalated to a human agent.

A confidence score represents the model's estimated probability that a prediction, classification, or response is accurate. Organisations set operational thresholds so high-confidence outputs are processed automatically, while lower-confidence results are routed to human review or a predefined fallback workflow.

One customer interaction received through a service channel. A single customer issue may generate several contacts if the person needs to call, email or message more than once.

A service operation that handles customer communication across multiple channels. Contact centres may manage telephone, email, chat, messaging, social media and other forms of support within the same operation.

A cloud-based contact-centre platform provided through a subscription service. Contact Centre as a Service, or CCaaS, may include routing, workforce management, reporting and support across several channels.

The underlying event, motivation or problem that caused the customer to seek assistance. Contact drivers may include failed payments, late deliveries, unclear bills, product faults or confusing website instructions.

Tracking contact drivers can reveal recurrent customer motivations and aide root cause analysis of recurrent problems.

The volume of customer contacts compared with a relevant business base. An organisation might measure contacts per customer, per order or per active account.

Contact rate measures the proportion of customers or transactions that result in a support enquiry. For example, a 7% rate equates to 70 customer contacts for every 1,000 orders processed or active users.

The recorded explanation for why a customer contacted the organisation. Contact reasons are commonly used for reporting, routing, workforce planning and identifying preventable demand or an underlying issue.

The number of customer interactions received during a selected period. Volume can be reported by channel, time of day, customer type, product or issue category.

The completion of a conversation or interaction within an automated service environment without a transfer to an employee. A good containment measure should be considered alongside resolution, repeat contact and customer feedback.

Information that helps a person or system interpret the customer's situation or issue almost immediately. This may include the current customer's conversation, previous contacts, recent transactions or interactions and related notes, account details or known service problems.

The amount of information a language model can process at one time. A larger context window allows more conversation history or source material to be included, although irrelevant information can still weaken the answer.

The practice of planning how a service conversation should work. It includes the language used, the order of questions, how errors are handled, when clarification is required and how the customer reaches a person when automation is no longer suitable.

Conversation flow is the sequence of messages, choices and actions that directs a chat from its trigger to its exit. It may follow a fixed scripted flow, adapt dynamically to customer input, or use a hybrid structure combining menus with open responses.

Nodes represent individual steps, while branches determine the different paths the conversation can take, including resolution or transfer to a human agent.

Technology that allows people to interact with software using ordinary written or spoken language. Conversational AI can appear in chatbots, voice systems, virtual assistants and messaging applications.

An AI tool designed to work beside an employee. In customer service, a copilot may locate information, draft a response or summarise the case, while the employee retains control over the final action.

The average cost incurred when handling one customer interaction. The calculation may include employee time, technology, facilities, supplier charges and management overheads, depending on how the organisation has defined it.

CCS measures how strongly customers trust a business to act in their best interests, keep its promises and provide a consistently reliable experience. It is commonly measured using a rating scale, such as 1 to 10, and can provide an indication of long-term customer trust and loyalty.

A measure of how easy or difficult a customer found a task, journey or service interaction. Customer Effort Score, or CES, is often collected by asking customers to rate how much effort was required to resolve their issue, normally rated on a Likert scale from 1-5 or 1-7.

Customer Effort Score (CES) is used to assess how easy or difficult it is for a customer to complete a purchase, use a product or resolve a support issue. Lower effort is generally associated with stronger loyalty, while higher effort can increase the risk of customer churn.

The overall impression a person develops through their dealings with an organisation. Customer experience, commonly shortened to CX, is shaped by the combined effect of products, service, systems, communication and operational decisions.

A combined measure used to estimate the condition of a customer relationship. It may consider product use, support activity, satisfaction, payment history and other indicators of retention risk.

The path a customer follows while trying to achieve something with an organisation. A journey may begin before purchase and continue through delivery, product use, support, renewal or cancellation.

The process of documenting a customer journey so that the experience can be examined.

A journey map may show customer actions, service channels, expectations, delays, emotions and the internal processes supporting each stage.

An estimate of the value a customer is expected to generate during the full relationship with the organisation. Customer lifetime value is often abbreviated to CLV or LTV.

A customer's tendency to continue choosing an organisation over time. Loyalty may be reflected through repeat purchasing, renewal, recommendation or resistance to switching.

The process of helping a new customer begin using a product or service. It may include account setup, training, guidance, expectation setting and early support.

The combination of technology and working practices used to manage customer relationships. A customer relationship management platform, usually called a CRM, can hold contact details, account history, service records, sales activity and previous communications and be used to tailor outbound interactions based on information the company collects about a customer.

The organisation's ability to keep customers over a defined period. Retention is influenced by the value of the product, service quality, pricing, competition and the wider customer experience.

Customer Effort Score (CES) measures how easy or difficult customers find it to complete a transaction, use a product or resolve a support issue. It is closely linked to customer retention because repeated friction can reduce satisfaction and encourage customers to switch providers, while straightforward experiences make them more likely to remain with the organisation over time.

A measure of how pleased customers are with an interaction, journey, product or service. Customer satisfaction, commonly shortened to CSAT, is usually collected through a rating question. The reported score often represents the percentage of respondents who selected a positive response.

The assistance an organisation provides to people before, during and after they purchase or use a product or service.

The use of technology to complete service work that would otherwise require manual action. Automation may support an employee, complete one part of a process or manage the entire interaction.

The planning and oversight of service employees, processes, technology, standards and performance.

An employee who responds to customer enquiries and service requests. Customer service representative is often shortened to CSR. Other titles include customer care adviser, support specialist and service agent.

Customer stickiness describes how strongly customers are drawn to continue using a product or service because it is useful, familiar or embedded in their routines. High stickiness can support retention by reducing the likelihood that customers will switch to an alternative.

The perceived difficulty associated with switching to another service or product can also influence stickiness.

A business function focused on helping customers achieve ongoing value from a product or service. It is particularly common in subscription and software companies, where the team may support onboarding, adoption and renewal.

Refers to the help and assistance provided when customers have a question, or encounter problem or technical difficulty while using a product or service.

A method of concealing sensitive information so that it cannot be viewed or reproduced in its original form. It may be used to hide card numbers, identification details or personal data within logs, transcripts and AI systems.

Techniques used to reduce tension, calm an upset customer and move the interaction towards a constructive outcome.

A structured set of questions and rules that leads the user towards a decision or outcome. Decision trees are widely used in troubleshooting, scripted support and traditional chatbot design.

The movement of customer demand away from employee-assisted service and towards another route, usually self-service or automation. Deflection is valuable when the alternative route solves the customer's problem. Making human support harder to access without improving the alternative simply relocates the frustration.

A problem that prevents an order from being delivered as expected, such as damage, delay or an incorrect address.

The process of understanding, reducing and directing customer contact across available service channels.

A customer who gives a low Net Promoter Score and may be dissatisfied or unlikely to recommend the organisation.

Customer support delivered through online channels. It includes email, live chat, messaging applications, social media, portals and digital self-service.

Online tools that allow customers to find information or complete tasks without contacting an employee.

Questions used to understand the customer's situation, needs and desired outcome before proposing a solution.

A label applied at the end of a customer interaction to describe what happened. Disposition codes may record the contact reason, outcome, next step or whether further action is required.

The process of investigating and resolving a disagreement relating to a charge, transaction, decision or service outcome.

The written record of a customer interaction, including the issue, actions taken, commitments made and final outcome.

Information provided to customers when a service or system is unavailable, including updates, expected recovery times and alternative options.

A change in the behaviour or performance of an AI system over time. Drift can develop when products, policies, customer language or real-world data change after the system has been deployed.

The proportion of customers who leave a service journey, form, queue or conversation before completing it.

Customer assistance provided through email, usually for enquiries that do not require an immediate response.

A mathematical representation of information that captures aspects of its meaning or similarity. AI systems use embeddings to compare customer questions with documents, previous conversations or knowledge articles, allowing customer questions to be matched with relevant documents, knowledge articles or previous conversations, even when the wording is different.

The use of technology to estimate emotional signals from text, speech or behaviour. These systems may attempt to identify frustration or satisfaction, but emotional meaning can be shaped by context, culture and individual communication style.

The ability to recognise and respond appropriately to a customer's feelings, concerns and circumstances.

The overall experience employees have at work, including the systems, support and conditions that affect service delivery.

The complete handling of a customer issue from the initial contact through to its final outcome.

A customer request for information, clarification or assistance.

The number or proportion of customer enquiries received during a defined period.

A distinct piece of information identified within language or data. In a customer message, an entity may include a person, location, date, product name, account number or order reference.

The automated process of identifying and capturing relevant information from a message or document. For example, a system may extract an order number, product name and requested delivery date from a customer enquiry.

The movement of a customer issue to a person or team with additional expertise, authority or system access. Escalation may be necessary when a case is complex, sensitive, high risk or outside the first agent's permissions.

A documented structure showing when an issue should be escalated, who should receive it and what level of authority is required. It may also include response times and communication responsibilities.

The proportion of customer contacts or cases passed to a higher support level. A high rate may indicate insufficient frontline authority, gaps in training or unusually complex demand.

The structured testing of a system to determine how well it performs. AI used in customer service may be evaluated for accuracy, relevance, safety, tone, policy compliance and resolution quality.

A collection of examples used to assess the performance of an AI model or AI-enabled customer service system. It should reflect the language, customer issues, uncommon scenarios and edge cases the system is likely to encounter after deployment.

The process of clearly communicating what a customer can expect, including timeframes, limitations and likely outcomes.

The structured collection and use of customer feedback and service data to improve the overall customer experience.

Methods, systems or documentation that help people understand how an AI output or recommendation was produced. Explainability becomes particularly important when the system influences decisions affecting customers.

A shortened form of Explainable Artificial Intelligence, often written as XAI. It refers to approaches that help people understand the basis of an AI recommendation, prediction or decision.

Customer contact created because a product, service or process did not work correctly the first time.

A response or route used when an automated system cannot continue confidently. The system may ask the customer to rephrase, offer a menu of choices or transfer the conversation to an employee.

The proportion of automated conversations that reach a fallback response. A high rate may indicate weak intent recognition, missing knowledge or poorly designed conversation paths.

Information provided by customers about their experiences, opinions or level of satisfaction.

The process of collecting, reviewing, responding to and acting on customer feedback.

Additional training applied to an existing AI model using a selected collection of examples. Fine-tuning may help a model follow a particular format, language style or specialised task more consistently.

The percentage of customer issues completed during the first interaction, without the customer needing to initiate another contact, or the company needing to schedule an outbound contact, about the same problem. First Contact Resolution is commonly shortened to FCR. Organisations need a clear definition of what counts as a repeat contact and how long the measurement window remains open.

The period between a customer sending an enquiry and receiving the first meaningful reply. First response time, or FRT, is frequently used for email, messaging and ticket-based support. Most companies aim for a first response within 24 hours of s customer contacting the business.

Additional contact made after an initial interaction to provide an update, confirm resolution or check the customer's experience.

The proportion of customer cases that require additional contact after the first interaction.

The degree to which predicted service demand matches the number of contacts that actually arrive. Accurate forecasting helps organisations schedule enough staff without creating unnecessary spare capacity or being under capacity.

The process of estimating future customer contact volumes to support staffing and resource planning.

Anything that makes a service journey harder than it needs to be. Repeated authentication, missing information, confusing forms and unnecessary transfers are common sources of friction.

The first level of customer service responsible for handling common enquiries and resolving routine issues.

The process of completing a customer order or service request, including preparation, delivery and confirmation.

A comparison between current service performance and the standard or outcome the organisation aims to achieve.

A broad customer question that does not relate to a complaint, technical issue or specific transaction.

AI that produces new content in response to prompts, instructions or contextual information. In customer service, it can draft replies, summarise conversations, explain policies, personalise messages and generate suggested responses for agents. Its outputs are created from learned patterns and supplied context, so they may require grounding, validation or human review before use.

Compensation or another benefit offered to recognise inconvenience and help restore the customer relationship.

The opening message or statement used to welcome a customer and begin an interaction.

Connecting an AI response to relevant and verifiable information rather than relying solely on patterns learned during model training. Grounding sources may include company policies, current product information, authorised customer records, knowledge articles or trusted external data.

A control introduced to reduce unsafe, inaccurate or unauthorised AI behaviour. Guardrails can restrict actions, block certain content, require verification or direct high-risk requests to an employee.

Customer service provided to people using a service without a registered account or established customer profile.

Assistance that leads a customer through a process using structured instructions, prompts or recommended steps.

An AI-generated statement that appears credible but is unsupported, invented or wrong. Hallucinations are especially risky when the system provides policy, financial or account-specific information.

The operational time used to complete a customer interaction and its related administration.

The transfer of responsibility for a customer conversation from one person or system to another. A well-designed automated handoff carries forward the customer's messages, relevant account details and the reason for transfer.

An online area where customers can search for support information. It may contain policy explanations, troubleshooting instructions, frequently asked questions and contact options.

The amount of time a caller spends waiting after an employee has answered and placed the call on hold.

The wider controls through which employees supervise automated systems. Oversight may include monitoring, quality review, incident handling and the ability to stop or override the system.

A working arrangement in which a person reviews, approves or contributes to an AI-supported decision or action.

The process of confirming that a person is authorised to access an account, receive information or request a change.

The outcome a customer is trying to achieve through their message. Examples include reporting a fault, tracking an order or changing a booking.

The process through which a conversational system estimates what the customer wants. The recognised intent can determine the next question, answer or workflow.

Any exchange between a customer and an organisation. Interactions may involve employees, automated systems or a combination of both.

The examination of customer conversations and service records to identify patterns. It may be used to study contact drivers, recurring problems, sentiment, compliance and resolution outcomes.

An automated telephone system that guides callers through spoken instructions or keypad choices. Interactive voice response is usually shortened to IVR.

The analysis of customer behaviour across several stages, interactions or channels. Unlike a single-contact measure, journey analytics looks at how the experience develops over time.

The coordination of service actions, communications and channels according to the customer's current circumstances.

A written resource explaining a product, policy, process or solution. Knowledge articles may be written for customers, employees or both.

An organised collection of information that can be searched by customers, service employees or automated systems.

A subject that is absent, outdated or insufficiently explained within a knowledge source. Knowledge gaps often become visible through repeated contacts, failed searches or recurring agent questions.

The process of creating, maintaining, organising and distributing useful information across an organisation.

An approach in which service employees create, improve and reuse knowledge while resolving customer issues. Knowledge-Centred Service is commonly shortened to KCS.

An AI model trained on extensive language data so that it can interpret and generate text. Large language models, commonly called LLMs, can support conversation, search, classification and summarisation. Their responses still require controls because fluent wording does not guarantee accuracy.

The delay between a user making a request and the system beginning or completing its response. High latency can make chatbots, agent-assist tools and voice systems feel unresponsive.

A digital service channel that connects a customer with an employee through text. The conversation may happen immediately or continue asynchronously, depending on the platform.

A field of computing in which systems identify patterns from data and use those patterns to produce predictions or classifications.

The average period required to resolve an issue. Mean Time to Resolution, or MTTR, is widely used in technical support, service management and incident reporting.

A computational system trained or designed to generate an output from information. In customer service, a model might predict churn, classify contact reasons or produce text.

A change in model performance caused by differences between current conditions and the information on which the model was originally developed.

The policies, responsibilities and controls used to manage AI models across their lifecycle. It may cover approval, documentation, testing, monitoring, access and retirement.

The provision of service through several channels that may operate separately. Customers have a choice of routes, but their information and conversation history may not transfer smoothly between them.

Technology used to produce written or spoken language from data or system instructions. Natural Language Generation is usually abbreviated to NLG.

The field concerned with how computer systems analyse, interpret and generate human language. Natural Language Processing is commonly shortened to NLP.

The part of language technology concerned with identifying meaning, context, intent and important details within human communication. Natural Language Understanding is often abbreviated to NLU.

A customer advocacy metric based on how likely someone says they are to recommend an organisation, product or service. Responses are grouped as Promoters (9–10), Passives (7–8) and Detractors (0–6). NPS is calculated by subtracting the percentage of Detractors from the percentage of Promoters and is reported as a score from -100 to +100.

A recommended action based on available customer information, predictive analysis or business rules. The recommendation may involve a service response, an offer, an escalation or a follow-up.

The process of responding to a customer's concerns, doubts or resistance. It involves understanding the reason behind the objection and providing relevant information rather than simply attempting to overcome it.

The proportion of available employee time spent handling customer work or completing related administration.

A service model in which channels, customer information and conversation history are connected. The aim is to allow customers to move between service routes without starting again.

Customer contact initiated by the organisation. Examples include callbacks, service updates, renewal reminders and proactive notifications.

The adjustment of a service experience according to the customer's circumstances, history or preferences. Personalisation can improve relevance, although it depends on accurate information and appropriate privacy controls.

Information that can identify a person directly or when combined with other information. Personally identifiable information, or PII, may include names, addresses, account details, identification numbers and contact information.

The use of historical data and statistical methods to estimate future outcomes. Customer service teams may use predictive analytics to forecast contact demand, identify customers at risk of leaving or anticipate escalation.

The use of data and predictive models to select the agent or service route most likely to produce a successful outcome for a particular customer.

Support initiated before the customer asks for help. Examples include notifying someone of a delivery delay or warning them about a likely account problem.

The instruction, question or contextual information given to a generative AI model.

The design and refinement of instructions intended to produce a more useful or controlled AI response.

An attempt to alter an AI system's behaviour through instructions hidden or included within user content, documents or external data. Without suitable safeguards, prompt injection may cause a system to disregard its intended rules or reveal information it should not provide.

The structured review of customer interactions against defined standards. Quality assurance, usually shortened to QA, may assess accuracy, communication, process adherence, compliance and whether the issue was handled appropriately.

The ongoing examination of service interactions to identify strengths, risks and coaching needs.

A collection, normally sorted chronologically, of customer contacts waiting to be handled. Queues may be organised by language, channel, product, skill or customer type.

The amount of time a customer waits before their interaction is accepted.

The comparison of an employee's current activity with their planned schedule. It helps workforce teams identify unexpected gaps in coverage while service demand is occurring.

The analysis of service activity while a conversation or event is still taking place. Real-time analytics can alert supervisors to risk, sudden demand or a customer who may need additional support.

The deliberate testing of an AI system to uncover vulnerabilities, unsafe responses or ways of bypassing its controls.

The proportion of closed or resolved tickets that later return to an open status. A high reopen rate may suggest that issues are being closed before the customer's need has been fully addressed.

The proportion of customers who contact the organisation again about the same or a connected issue.

The point at which the customer's request or problem has been dealt with.

The proportion of contacts or cases that reach a completed outcome.

The period between a customer message and the organisation's reply.

The percentage of customers who remain with an organisation over a selected period.

An AI method that searches an approved information source before creating a response. Retrieval-Augmented Generation, commonly called RAG, can connect a language model to policies, product material or knowledge articles. The response still depends on whether the correct information was retrieved and interpreted properly.

The investigation of the underlying reason a customer problem occurred. It aims to address the source of repeated failures rather than resolving each contact separately.

The process of directing a customer interaction to the most appropriate queue, employee, team or automated process.

A chatbot that follows predetermined rules, menus or decision paths. Rules-based bots can work reliably within a narrow area but may struggle when customers phrase requests in unexpected ways.

The extent to which an employee follows their planned working schedule. It may include login times, breaks, meetings and periods assigned to customer handling.

A service route that allows customers to find information or complete a task without direct employee assistance.

The automated classification of emotional tone within text or speech. Sentiment systems often use categories such as positive, neutral or negative, although they can miss sarcasm, mixed emotions and cultural differences.

A central support function that receives and manages requests, faults and access issues. The term is most often associated with IT support but may also apply to internal employee services.

A measure showing how quickly a defined proportion of customer contacts are answered. A contact centre may report, for example, the percentage of calls answered within a particular number of seconds.

The actions taken after something has gone wrong with the service. The aim is to resolve the immediate problem and reduce the damage caused to the customer relationship.

A documented commitment setting out an expected service standard. A service-level agreement, or SLA, may cover response times, resolution times, system availability or supplier performance.

The proportion of paid employee time that cannot be used for handling customer demand. It may include leave, training, meetings, breaks, system outages and other non-contact activities.

A combined record of customer information gathered from different systems and interactions. A Single Customer View, sometimes called Customer 360, helps employees understand the customer's history without searching across several platforms.

The assignment of customer contacts to agents whose skills match the enquiry. Routing may consider product knowledge, language, technical expertise or authority level.

Technology used to examine recorded or live telephone conversations. It may identify contact reasons, repeated phrases, silence, interruptions, sentiment or possible compliance concerns.

Technology that converts spoken language into written words.

The use of AI to produce a condensed record of a conversation, ticket or case.

The period during which the employee and customer are actively speaking on a call.

Technology that turns written text into spoken audio. Text-to-speech is commonly abbreviated to TTS.

A digital record created to track a customer request, question or problem.

The amount of time a support ticket has remained open.

The avoidance of a support ticket because the customer completed the task or found the answer through another channel. A deflected ticket should represent a successful alternative journey rather than a customer who gave up trying to reach support.

The process of recording, prioritising, assigning and resolving customer tickets.

A classification showing how urgently a support ticket should be handled. Priority may be determined by customer impact, risk, severity or the number of people affected.

A label showing the current stage of a customer ticket, such as new, open, pending, resolved or closed. Statuses help teams track ownership and outstanding work.

A service structure in which cases are divided according to complexity or required expertise. Routine enquiries may remain with frontline service, while specialist or technical cases move to higher tiers.

A unit of language processed by a large language model. A token may be a complete word, part of a word or punctuation. AI usage limits and costs are often calculated using token volume.

The information used to teach an AI system how to recognise patterns or produce outputs.

The percentage of contacts moved from one employee, team, queue or system to another.

The initial assessment of an incoming contact to determine its urgency, subject and appropriate destination.

One contribution within a conversation. A customer message forms one turn, followed by another turn when the employee or system replies.

A customer interaction that ends without the request being completed or the problem being adequately addressed.

The percentage of time a service or system remains available and operational.

A single piece of language used by a customer during a conversation. In chatbot design, an utterance may be an example phrase associated with a particular intent.

A database designed to store and search mathematical representations of information. Vector databases are commonly used in AI retrieval systems to locate documents with meaning related to the customer's question.

An automated system that communicates with customers and completes service activities. The term is often used for systems with broader capabilities than a simple scripted chatbot.

A conversational system that helps users find information, complete tasks or access services.

Artificial intelligence used within spoken customer interactions. Voice AI may combine speech recognition, language processing, system integrations and generated speech.

The collection and analysis of what customers say, expect and experience. Voice of the Customer, commonly shortened to VoC, may draw on complaints, surveys, reviews, interviews and conversation data.

An automated system that communicates with customers by voice.

The period a customer spends waiting for assistance, a response or a completed outcome.

A transfer in which the current agent speaks with the receiving employee before connecting the customer. Relevant details can be passed on so the customer does not need to explain the issue again.

A mechanism that allows one system to send information to another as soon as a specified event occurs. For example, a support platform may send a webhook when a new ticket is created or its status changes.

A call function that allows a supervisor to speak privately to an agent during a live interaction without the customer hearing the guidance.

A defined sequence of tasks and decisions used to complete a process.

The use of technology to perform or coordinate steps within a workflow. A customer service workflow might classify an enquiry, retrieve account data and send the case for approval before confirming the result.

The forecasting, scheduling and staffing processes used to match employee capacity with customer demand. Workforce management is commonly shortened to WFM.

The time used after a customer interaction to complete notes, categorisation and administration. Wrap-up time is another name for after-call work.

The ability of an AI model to carry out a task without having been trained on examples created specifically for that exact task.

The completion of a customer request through automation without direct employee involvement. A zero-touch outcome should still be assessed for accuracy, customer effort and whether the customer's actual need was resolved.