Glossary (188)

A framework that allows AI agents to communicate, collaborate, negotiate, and exchange information directly with one another.

An open protocol designed to enable communication, coordination, and task delegation between AI agents operating across different systems, platforms, or organisations. ACP provides a standardised framework for agents to exchange instructions, context, capabilities, and results while maintaining interoperability and governance.

Testing AI with difficult, malicious, or edge-case inputs. Example: a guest attempts to trick a chatbot into providing unauthorised discounts, confidential information, or policy exceptions.

An AI system capable of autonomously completing tasks, making decisions, and interacting with software, data, or users.

Stored context that allows an AI agent to remember preferences, previous actions, or operational patterns. Useful, but sensitive from a privacy and governance perspective.

A modular capability added to an AI agent, such as accessing files, sending emails, browsing the web, querying databases, or interacting with external systems.

A group of AI agents collaborating toward a shared objective while performing specialised tasks.

A commercial model in which AI agents negotiate and transact on behalf of users or businesses.

A future distribution model in which travel planning, shopping, booking, and post-booking interactions increasingly occur between autonomous agents rather than between humans and websites.

AI designed to plan, reason, and act toward goals with minimal human intervention.

The criteria and algorithms that determine how autonomous AI agents evaluate, filter, and position hotel offers when presenting recommendations to users.

The application of autonomous AI agents to continuously analyse data, make pricing and inventory decisions, execute revenue strategies, and optimise commercial performance with minimal human intervention.

A digital marketplace where AI agents can discover, purchase, exchange, or access services, skills, data, and capabilities from other AI agents.

The idea that conversational AI becomes the primary interface through which users interact with software, information, and digital services. Rather than navigating menus, forms, dashboards, or websites, users communicate through natural language, with AI acting as the intermediary layer between human intent and system execution.

An AI-powered conversational system that supports users through natural language interactions.

An AI assistant that provides recommendations, answers questions, and facilitates guest services.

The policies, controls, and principles governing the responsible use of AI.

Boundaries and controls that prevent AI from exposing guest data, making unauthorised promises, changing rates without approval, sending unapproved communications, or generating unsafe recommendations.

A confident but incorrect answer generated by AI.

The superficial application of AI tools across a legacy technology environment that lacks the foundational data structures or integration necessary to support meaningful outcomes.

A basic understanding of what AI can do, what it cannot do, and how to use it responsibly. A concept strongly promoted by Ian Millar.

AI-generated meeting minutes including key decisions, action items, owners, and deadlines from operational or corporate meetings.

The coordination of multiple AI models, agents, systems, tools, and workflows toward a common objective.

AI-generated summaries displayed directly within search results.

A formal document defining how employees may and may not use AI within an organisation.

An evaluation of whether a hotel possesses the data, systems, governance, skills, workflows, and leadership support required for successful AI adoption.

A step-by-step plan outlining AI adoption priorities, timelines, and implementation phases across departments.

The measurable return generated by AI adoption, including time savings, revenue growth, cost reduction, improved guest satisfaction, reduced risk, or increased productivity.

The modern paradigm of search functionality, where users receive synthesised, conversational responses that invite ongoing dialogue, replacing the static list of links provided by traditional search engines.

Dependence on a single AI vendor’s models, infrastructure, integrations, data structures, or agent ecosystem.

Hospitality systems designed with embedded AI capabilities from inception, rather than having AI features added later.

Clean, structured, current, accurate, and trusted data that AI systems can effectively use.

Concept introduced by Edmondo Grassi, describing a social system in which agency, decision-making, and subjectivity emerge from the continuous interaction of humans, algorithms, animals, technical objects, and environmental entities, forming a distributed, co-evolutionary technoecology rather than a human-centered hierarchy.

One of the three hospitality models within the Hospitality Evolution Framework. In the Anthropocentric model, human employees remain the primary source of value creation and service delivery. Human interaction, empathy, judgement, and presence are positioned as premium features, and guests may be willing to pay more for a predominantly human experience.

A protocol designed to facilitate financial transactions between autonomous AI agents. AP2 enables agents to negotiate, authorise, and execute payments on behalf of users or organisations, supporting emerging models of agentic commerce where software agents can independently purchase services, settle transactions, and manage economic interactions.

Data updated periodically rather than continuously in real time. Less suitable for time-sensitive operational decisions.

Systematic distortion in data or AI outputs.

An AI system whose reasoning is difficult to explain.

Digital content structured and optimised primarily for consumption, interpretation, and processing by autonomous AI agents rather than human readers.

The process of ensuring AI-generated communications reflect the hotel’s approved tone, style, and brand personality rather than sounding generic or machine-generated.

Enterprise-grade, industry-specific AI tools or native AI capabilities embedded within business applications. Unlike Consumer AI, the adoption, deployment, and configuration of Business AI are entirely within the control of the business or hospitality operator.

Software that interacts through text or voice conversations.

A conversational AI application developed by OpenAI that brought generative AI into mainstream public awareness. Commonly, the first AI tool hotel employees encounter independently, making it a frequent entry point for Shadow AI (see).

The process of breaking down large datasets or documents into smaller, semantically coherent segments to optimise data retrieval in AI systems like RAG.

A conversational AI assistant developed by Anthropic. Known for handling long documents and complex instructions. Used in hospitality for analysing contracts, drafting SOPs, summarising guest feedback, and powering AI agents. Many hotel-facing tools may be built on Claude without the hotelier being aware of it.

An AI-powered coding tool developed by Anthropic that allows users to build, edit, and debug software through natural language instructions. Relevant to hospitality operators building internal tools or automations without large engineering teams.

An enterprise AI environment configured to protect organisational data and prevent it from being used to train public AI models.

A platform used to collect, manage, and document user consent.

A metric indicating the level of certainty an AI model has in its generated response or decision, often used to mitigate the risk of hallucinations.

A bridge between AI systems and data sources or operational platforms, such as PMS, POS, RMS, BMS, CRM, or document repositories.

Publicly available, mainstream AI tools adopted independently by the general public. In a hospitality context, this refers to the technology the guest chooses to use, entirely outside the hotel's or brand's control or influence.

The amount of information an AI model can consider at one time. Determines how much content, such as contracts, SOPs, guest histories, incident reports, or revenue packs, can be analysed in a single interaction.

Users interact with analytics through natural language rather than dashboards.

An AI assistant working alongside a human user.

AI analysis of the true cost of a booking, including commissions, transaction fees, loyalty costs, marketing spend, and distribution expenses.

A framework for managing hotel data in the age of AI and agentic commerce. CUPS outlines four essential steps to make hospitality data discoverable, trustworthy, and actionable for both humans and machines: consolidating data into a single source of truth, keeping it up to date, processing it into machine-readable formats, and sharing it across distribution channels and AI agents.

An AI-powered code editor closely associated with the Vibe Coding movement. Increasingly used by hospitality operators to build lightweight internal tools, automations, and data workflows without relying on dedicated engineering resources.

The policies, processes, and controls that ensure data is accurate, secure, compliant, and properly managed so it can be trusted for business operations, analytics, and AI.

A centralised repository for large volumes of data.

Information trapped within separate systems such as PMS, POS, RMS, CRM, GRMS, BMS, reputation platforms, HR systems, or finance tools.

A machine learning technique based on neural networks.

An AI capability that autonomously performs multi-step research, gathering, analysing, synthesising, and validating information from multiple sources to produce comprehensive outputs.

AI that predicts future demand using booking pace, market conditions, events, competitor activity, weather, flight schedules, search behaviour, and historical trends.

Work performed by AI systems, agents, or robots that substitutes or augments human labour.

A paradigm shift where the direct booking engine serves primarily as an API or execution environment for AI agents to negotiate and finalise transactions autonomously.

AI running directly on local devices.

Numerical representations of text, images, or other data that capture their meaning, allowing AI systems to compare, search, and retrieve information based on semantic similarity rather than exact keywords.

The ability to search across an organisation’s internal documents, systems, and knowledge repositories using AI.

Rules defining when AI must transfer a conversation or task to a human. Examples include angry guests, VIP complaints, refund disputes, legal threats, safety incidents, medical concerns, harassment reports, or payment disputes.

A traveller who considers the ethical, social, environmental, and technological implications of their decisions and behaviours throughout the travel journey.

The European Union’s regulation governing the development, deployment, and use of AI systems, establishing obligations based on the level of risk posed by each application.

Adapting a foundation model for a specific task or industry.

Frontier models are the most advanced AI models available at a given point in time, defining the current leading edge of what artificial intelligence can do.

A structured method allowing AI systems to trigger predefined actions or workflows, such as creating maintenance tickets, retrieving policies, or summarizing arrivals.

A family of AI models developed by Google, integrated across Gmail, Google Docs, Google Search, and Google Cloud. Relevant to hoteliers because it powers AI features within tools that many properties already use.

AI capable of creating text, images, audio, video, or code.

The practice of optimising content to increase visibility within AI-generated responses. See AEO.

An AI coding assistant developed by GitHub that helps developers write code faster. Commonly used by the technology vendors and integration partners building hotel systems.

Concept introduced by Mark Fisher. A liminal condition in which the distinction between human and machine, animate and inanimate, becomes increasingly blurred. It serves as a metaphor for the technological transition currently reshaping hospitality.

The family of AI models developed by OpenAI that powers ChatGPT and many third-party applications. Many hotel-facing AI tools and chatbots are built on GPT models, sometimes without this being explicitly stated by the vendor.

AI identification of guests who may not return due to dissatisfaction, reduced engagement, price sensitivity, poor service recovery, or changing behaviours.

A digital representation of a guest’s preferences, behaviours, and history.

Robotics-based artificial intelligence that integrates cognitive software with physical hardware, enabling machines to perceive, navigate, and interact with the physical world.

A content management system that separates content storage from presentation.

An AI application involving sensitive decisions, guest data, pricing, employment, health, safety, legal exposure, or brand reputation. Requires stronger controls and human approval.

A framework describing the evolution of hospitality through three operating models: Anthropocentric, Hybrid, and Technocentric, based on the relative contribution of human and digital workers to value creation and service delivery.

A proposed ethical framework for hospitality professionals, inspired by the Hippocratic Oath in medicine. It represents a commitment to preserving the sacredness of human-to-human relationships and to ensuring that technology enhances rather than replaces genuine hospitality.

A state of auto-automation guided by human oversight, where human captains provide direction based on system feedback and contextual nuance. The goal is to maximise human-centred hospitality, strategically enhanced and given a competitive advantage through intelligent technology.

The ability for employees to stop, modify, reject, or correct an AI-generated action or recommendation. Essential when guest experience, safety, finances, or reputation are involved.

A strategic approach that measures technological success not only through efficiency gains but also through improvements in employee well-being, resilience, and quality of work.

The collection of human capabilities that become more valuable as AI adoption increases, including empathy, judgment, creativity, ethics, leadership, cultural awareness, and relationship-building.

Humans actively review or approve AI outputs.

Humans supervise AI systems and intervene when necessary.

AI operates without direct human supervision.

The theory that, as automation and artificial intelligence become increasingly ubiquitous, authentic human interaction, craftsmanship, empathy, judgement, presence, and even imperfection become scarcer and therefore more valuable, eventually evolving into a premium form of luxury. Original paper: “Humans-as-Luxury: The Future of Hospitality in an AI-Driven Age,” available at https://humansasluxury.com.

The strategic coordination of technology, employees, data, timing, service design, and human touchpoints to ensure guests feel recognised rather than processed.

One of the three hospitality models within the Hospitality Evolution Framework. In the Hybrid model, value creation is shared between human employees and intelligent systems. AI, automation, and robotics handle routine and transactional tasks, while humans focus on empathy, creativity, problem-solving, and guest experience. The model seeks to balance operational efficiency with meaningful human interaction and is likely to become the predominant operating model for most hotels in the foreseeable future.

The combination of AI, automation, and process optimisation technologies.

A software application that provides developers with the tools needed to write, test, and debug code in one place. AI-powered IDEs such as Cursor, Windsurf, and GitHub Copilot now incorporate generative AI to assist with coding through natural language, lowering the barrier for non-traditional developers to build operational tools.

A structured package of a traveler’s needs, preferences, constraints, and trip goals, assembled by a personal AI agent or chatbot. It functions like a “mini-RFP” that can be shared with a hotel, travel seller, or seller-side agent to help them generate more relevant, customized offers. Where no seller-side agent exists, the traveler’s agent may use the intent envelope to call the seller’s available tools or MCP server directly.

The small group of options selected by AI systems before a traveller ever sees a list of results. Properties that fail to enter this shortlist become effectively invisible, regardless of their traditional search rankings.

A framework describing three strategic responses to labour shortages: Import People (addressing labour shortages through immigration, international recruitment, or workforce mobility. Typically a short- to medium-term solution, but often associated with high social and political resistance); Substitute People (replacing human labour with automation, robotics, artificial intelligence, and self-service technologies. Generally, the most scalable response to labour shortages and often faces lower political resistance than large-scale immigration); Produce People (expanding the labour force through demographic growth, education, training, and workforce development policies. Usually, a long-term solution requiring substantial time before results become visible).

An attempt to bypass AI safety restrictions.

A structured network of entities and relationships.

The process of locating and retrieving relevant information from databases, documents, or knowledge bases for use by AI systems.

The theory that hospitality temporarily recreates the illusion that the world revolves around the guest. Drawing on Jacques Lacan’s concept of the stade du miroir (mirror stage), the framework suggests that exceptional hospitality restores a fleeting sense of centrality, recognition, and importance, making guests feel as though their desires, needs, and presence occupy the centre of the experience. In this view, hospitality functions as a carefully orchestrated mirror that reflects the guest’s idealised sense of self.

A type of AI model trained on vast amounts of text.

The time elapsed between a user submitting an input and an AI system returning a response. For back-office tasks such as summarising reports or drafting communications, latency is rarely critical. For guest-facing applications like chatbots, voice agents and in-room assistants, it is a functional requirement: delays of more than one to two seconds create friction and undermine the guest experience. When evaluating AI vendors for real-time deployments, latency should be tested under realistic load conditions, not just controlled demos.

An open weights large language model developed by Meta. Unlike closed models, Llama can be deployed within a private or on-premise environment, making it relevant to hoteliers concerned about data privacy or vendor lock-in.

A reference or source provided by an AI system to support its answer.

The practice of optimising content, structure, and data so that AI models can more effectively discover, understand, and reference information.

An AI agent operating on a local device or within a private environment rather than entirely in the public cloud. Often preferred by privacy-sensitive organisations.

Context retained across multiple sessions. Requires clear privacy rules, retention policies, consent mechanisms, and deletion procedures.

An AI task with limited guest, legal, financial, or operational risk. Examples include summarising documents, drafting training materials, creating meeting agendas, or preparing internal briefing notes.

AI systems that learn patterns from data.

An open standard that allows AI models to securely connect to external systems, tools, databases, and applications.

The collection of memories, data, preferences, and digital traces that continue to exist after the physical guest journey has ended.

Microsoft's AI assistant integrated across Outlook, Word, Excel, PowerPoint, and Teams. Particularly relevant to hotel corporate offices already operating within the Microsoft ecosystem.

The ability to route tasks between different AI models based on performance requirements, privacy considerations, cost, or use case.

An environment in which multiple AI agents collaborate while performing specialised roles. Examples include Revenue Agents, Marketing Agents, Finance Agents, and General Manager Agents working together.

AI capable of understanding text, images, audio, and video together.

The dilution or elimination of human-delivered service within the guest journey, resulting from a technology strategy that shifts operational tasks entirely onto the customer. Characterised by an over-reliance on self-service automation (e.g., mandatory kiosks, rigid chatbots, app-only interactions), where efficiency gains for the operator come at the expense of authentic hospitality and personalised human connection.

An AI-powered research tool developed by Google that allows users to upload documents and interact with them through natural language. Useful in hospitality for summarising contracts, analysing reports, or building internal knowledge bases that staff can query conversationally.

AI systems whose code, model weights, or both are publicly available, allowing users to inspect, modify, deploy, and build upon the technology.

The American AI research company behind the GPT model family and ChatGPT. A significant proportion of the AI tools and chatbots available in the hospitality market are built on OpenAI's models and infrastructure.

An AI-powered search engine that provides direct, cited answers rather than a list of links. Relevant to hoteliers monitoring how potential guests discover and evaluate properties, as platforms like Perplexity can influence travel research without directing users to hotel websites.

Artificial intelligence embedded in robots and machines that can sense, navigate, and interact with the physical world. Example: Service robots for delivery, cleaning, and operational support in hospitality environments.

Using data and AI to forecast future outcomes.

AI proactively identifies anomalies, opportunities, risks, and emerging trends.

AI-assisted forecasting of equipment failures before breakdowns occur. Common applications include HVAC systems, elevators, pumps, chillers, boilers, laundry equipment, kitchen equipment, guestrooms, and building infrastructure.

A secure collection of organisation-specific information that AI can search, including SOPs, brand standards, menus, room information, policies, emergency procedures, FAQs, and training materials.

An instruction given to an AI system.

The practice of designing prompts to improve AI performance.

A curated collection of approved prompts designed for specific hotel departments, workflows, or roles.

An alternative solution to common consensus mechanisms, driving efficiency by removing human interference and establishing the objective of the AI to fulfil its task in the most useful way.

A publicly available AI application intended for general users. Useful, but potentially risky if confidential hotel, guest, employee, or ownership information is entered.

An AI architecture combining language models with external knowledge sources.

An AI configuration that can retrieve, analyse, or summarise information but cannot make changes. Often, the safest starting point for AI adoption.

Current operational information, such as room status, guest requests, occupancy levels, rate changes, work orders, payments, or maintenance alerts.

An AI model optimised for complex problem-solving.

A browser-based coding environment with integrated AI capabilities. Associated with the Vibe Coding movement, allowing hospitality operators to build and deploy lightweight tools without dedicated engineering support.

The use of AI, analytics, forecasting, and pricing strategies to maximise revenue, profitability, and commercial performance across hospitality operations.

AI technique that reverses the traditional prompting process to improve output quality. Instead of attempting to write the perfect prompt from the start, the user begins with the desired outcome and works backward, or allows the AI to ask clarifying questions in order to define goals, constraints, and requirements before generating a response.

The inversion of Masahiro Mori’s Uncanny Valley theory. Rather than artificial systems becoming unsettling because they appear almost human, the Reverse Uncanny Valley (also known as the Olympia Effect) describes the phenomenon whereby humans become so optimised, scripted, emotionally regulated, and system-dependent that they themselves begin to appear artificial and uncanny. In hospitality, it refers to the progressive transformation of employees into highly standardised service performers whose behaviours, language, emotions, and interactions increasingly resemble those of machines.

AI that drafts responses to online reviews. Outputs should be reviewed by humans to avoid generic, inaccurate, legally risky, or tone-deaf responses.

Technology that automates repetitive, rule-based digital tasks.

Executing AI actions within a restricted environment to reduce operational, security, and compliance risks.

A standardised vocabulary of structured data added to the website code that helps search engines, AI agents, and bots accurately understand, categorise, and extract specific content.

Sovereign, Ethical, Lifestyle Filtering. An acronym created by Jonathan MacDonald.

A business-friendly data layer that defines the meaning of key hotel metrics and concepts across systems, helping AI understand terms such as room revenue, net ADR, occupied room, guest complaint, or available inventory.

A hotel’s strategic ability to manage and influence how AI models interpret its brand, value proposition, and offerings compared to third-party distribution channels.

Search based on meaning rather than keywords.

The use of AI tools by employees without approval, governance, security review, or management visibility.

Context retained during a single task or session. Generally safer for sensitive hospitality operations.

The officially approved location where correct and current information resides. AI systems should reference this source rather than outdated documents or disconnected files.

A compact AI language model designed to operate with lower computational requirements, often enabling faster, cheaper, and more private deployments than large language models.

Software-based artificial intelligence applications that operate entirely within digital environments, focusing on data processing, language generation, analysis, and digital user interfaces.

An AI or software layer responsible for coordinating other agents and validating their outputs.

The ability to replace, upgrade, or change AI components within an ecosystem without disrupting operational continuity.

Artificially generated data used for testing, simulation, or model training without exposing real guest information.

The ability of AI-generated visual content to simulate human perception and attention, influencing how audiences interpret authenticity, trustworthiness, and reality.

The use of artificial intelligence to generate, personalise, and optimise persuasive messages in real time based on individual preferences, behaviours, and contextual signals.

The average generative AI cost incurred per guest. TCPG is calculated by dividing the total cost of AI tokens consumed across all guest-facing and operational interactions by the total number of guests served.

One of the three hospitality models within the Hospitality Evolution Framework. In the Technocentric model, AI, automation, robotics, and autonomous systems become the primary source of value creation and service delivery. Human involvement is minimised and reserved for exceptional situations, while guests increasingly interact with digital systems and self-service technologies. In this model, hospitality becomes increasingly defined by system performance rather than human interaction.

The basic unit processed by a language model.

The practice of excessively increasing AI usage, context length, agent complexity, prompt size, or workflow automation in pursuit of marginal performance gains, often resulting in disproportionate token consumption, rising costs, increased system complexity, and diminishing operational returns.

The full cost of AI adoption, including licenses, integrations, data preparation, training, governance, support, vendor fees, and change management.

Information used to train or improve an AI system. In hospitality, this may include SOPs, guest interactions, service logs, reviews, maintenance records, call transcripts, menus, or booking data.

The architecture powering most modern AI models.

A future state in which artificial intelligence, robotics, autonomous systems, and algorithmic decision-making become so deeply embedded throughout the travel ecosystem that most operational, commercial, and experiential processes occur with minimal human intervention, creating a point beyond which travel can no longer be meaningfully understood or operated through exclusively human decision-making.

The ability to recognise and understand guests as unique individuals rather than as collections of transactional or behavioural data points.

An open standard developed by Google designed to enable agentic commerce across AI surfaces such as Google Search's AI Mode and Gemini. UCP allows merchants, including hotels, to integrate checkout logic directly with Google's AI systems, enabling autonomous agents to complete bookings and transactions on behalf of users. Notably expanding into the lodging sector, UCP represents a significant shift in how hotels may need to think about distribution, as transactions increasingly bypass traditional booking interfaces entirely. Hotels that adopt UCP retain their customer data and remain the Merchant of Record.

Information such as emails, reviews, images, videos, and documents.

A database optimised for storing and searching embeddings.

A software development approach in which developers describe outcomes in natural language, and AI generates much of the code.

An AI-powered digital employee or assistant.

An autonomous AI agent that interacts through spoken conversation and can perform actions on behalf of users.

AI systems that understand and generate spoken language.

The purchase of products or services through voice interactions.

International standards defining how digital experiences should be made accessible.

An implementation of MCP designed for web-based systems and applications.

An AI-powered code editor developed by Codeium, similar in concept to Cursor. Part of the broader Vibe Coding movement, enabling non-traditional developers to build operational tools and automations.

The automation of repetitive business processes.

A permission level allowing AI to create, edit, send, approve, publish, or modify records. High risk and typically subject to strong governance controls.

The application of zero-trust security principles to AI systems and agents, where no user, system, or action is automatically trusted, and every request must be verified.

A search experience in which users receive answers directly without visiting external websites.

The ability of an AI model to perform a task without prior task-specific training.