Glossary · 60 terms
The language ofAI implementation,defined.
Plain-English definitions of the terms your board, your vendors and your team use about AI, agents and their governance. Each term opens to a full definition, what it means in practice, related terms and sources.
60 terms
A
- Strategy & adoptionAgent washingMarketing existing chatbots, assistants or automation tools as 'AI agents' without real agentic capabilities.
- Agents & automationAgent-to-agent protocol (A2A)An open protocol that lets AI agents built by different vendors discover each other, exchange tasks and share results securely.
- Agents & automationAgentic AIAI systems that pursue a goal by planning and taking multi-step actions, such as calling tools and checking results, rather than only answering a question.
- Commerce & industryAgentic commerceAI agents that research, compare and buy products on a shopper's behalf, and the standards and storefront changes that let them do it.
- Agents & automationAgentic workflowA business process in which AI models carry out several of the steps, following a path defined in code, with people handling direction and exceptions.
- Operations & costAgentOpsThe practices and tools for running AI agents in production: deploying, monitoring, evaluating, securing and improving them over time.
- Agents & automationAI agentA piece of software in which a language model works in a loop: it chooses actions, uses tools, observes the results and continues until a task is complete.
- Strategy & adoptionAI centre of excellenceAlso: AI CoEA central team that sets AI standards, shares expertise and tools, and helps business units find, build and scale AI use cases.
- Governance & regulationAI governanceThe policies, roles, processes and controls an organisation uses to decide which AI to build or buy and keep its use lawful, safe and aligned with its goals.
- Strategy & adoptionAI literacyThe skills, knowledge and understanding people need to use AI effectively and safely, and to judge its outputs, risks and limits.
C
- ModelsChain of thoughtA model working through intermediate reasoning steps before giving its answer, which improves accuracy on multi-step problems.
- Strategy & adoptionChange managementHelping people move from the old way of working to the new one, so that an AI system is actually adopted and its benefits realised.
- Building blocksComposite AICombining different AI techniques (machine learning, language models, rules, knowledge graphs, optimisation) in one solution, each used where it works best.
- Agents & automationComputer useAI agents that operate a computer the way a person does, by reading the screen and moving the mouse and keyboard, to work in software that has no API.
- Building blocksContext engineeringDeciding what information goes into a model's limited context window at each step (instructions, documents, tool results, memory) so it can do the task well.
- ModelsContext windowThe maximum amount of text, measured in tokens, that a model can consider at once, including the instructions, documents, conversation and its own answer.
- Agents & automationCopilotsAI assistants built into everyday software that help a person work faster (drafting, summarising, answering, coding) while the person stays in control.
D
E
- Building blocksEmbeddingsLists of numbers that represent the meaning of a piece of text, an image or another item, so that similar meanings sit close together.
- Governance & regulationEU AI ActThe European Union's law on artificial intelligence, Regulation (EU) 2024/1689, which sets rules by risk level for AI placed on or used in the EU market.
- Quality & safetyEvalsAlso: AI evaluationsRepeatable tests that measure whether an AI system does its job well, run before launch and on every change to prompts, models or data.
- Governance & regulationExplainabilityThe ability to give people a meaningful account of why an AI system produced a particular output or decision.
F
- Building blocksFine-tuningFurther training an existing model on your own examples so it adopts a format, tone or narrow skill more reliably than prompting alone achieves.
- Operations & costFinOps for AIApplying cloud financial management practices to AI: making AI spend visible, allocating it to teams and use cases, and optimising it against business value.
- Strategy & adoptionForward-deployed engineeringAlso: FDEEngineers embedded with a customer, working inside its systems and teams, to get AI working on real problems in production.
- ModelsFoundation modelA large model trained on broad data that can be adapted to many tasks, such as a large language model or an image model, rather than built for one purpose.
G
- Quality & safetyGroundingTying a model's output to trusted sources, such as your documents, databases or live search, so answers reflect real data rather than the model's memory.
- Quality & safetyGuardrailsPolicy and technical limits on what an AI system can say or do, such as input and output filters, permission checks and rules for when to stop.
H
- Quality & safetyHallucinationAlso: confabulationWhen a model produces fluent, confident output that is false or unsupported, such as an invented fact, citation, number or policy.
- Governance & regulationHuman-in-the-loopDesigning an AI system so a person reviews, approves or corrects its work at set points, especially for decisions that are high-impact or uncertain.
I
- Operations & costInference costWhat it costs to run a model each time it is used, driven by tokens processed, model size, reasoning effort and how the model is hosted.
- Governance & regulationISO/IEC 42001The international standard for an AI management system: requirements for how an organisation governs, develops, provides and uses AI responsibly.
L
- ModelsLarge language model (LLM)A neural network trained on vast amounts of text to predict the next token, giving it broad abilities to write, summarise, reason over and transform language.
- Operations & costLLMOpsThe operational practices for applications built on large language models: prompt and model versioning, evaluation, deployment, monitoring and cost control.
- Commerce & industryllms.txtA proposed plain-text file at a website's root that gives language models a short, structured guide to the site's most important content.
M
- Operations & costMLOpsThe practices and tooling for reliably building, deploying and maintaining machine learning models in production, borrowing from DevOps.
- Agents & automationModel Context Protocol (MCP)An open standard for connecting AI applications to tools and data sources, so one connector works with any model or app that supports the protocol.
- ModelsModel distillationTraining a smaller 'student' model to imitate a larger 'teacher' model, to get similar quality on a task at a fraction of the cost and latency.
- Governance & regulationModel riskThe risk of loss or harm from decisions based on models that are wrong, misused or used outside their intended purpose.
- Agents & automationMulti-agent orchestrationCoordinating several AI agents, each with its own role and tools, so that together they complete a task too large or varied for one agent.
N
O
- Operations & costObservabilityCollecting traces, logs and metrics from AI systems (every prompt, tool call, cost and outcome) so teams can see what happened and why.
- ModelsOpen-weight modelsModels whose trained parameters (weights) are published, so anyone can download, run and adapt them on their own infrastructure, within the licence terms.
P
- Commerce & industryPhysical AIAI that perceives and acts in the physical world through robots, vehicles, drones and machines, trained largely in simulation.
- Building blocksPrompt engineeringWriting and refining the instructions, examples and structure given to a model so it produces the output you need reliably.
- Quality & safetyPrompt injectionAn attack in which text crafted by an attacker, in a message, web page, email or document, tricks a model into ignoring its instructions or misusing its tools.
R
- ModelsReasoning modelsLanguage models trained to think through a problem step by step before answering, spending more compute at answer time to do better on hard tasks.
- Quality & safetyRed teamingDeliberately attacking an AI system, as an adversary would, to find harmful outputs, security holes and failure modes before real users or attackers do.
- Governance & regulationResponsible AIDeveloping and using AI in line with principles such as fairness, transparency, accountability, privacy, safety and respect for human rights.
- Building blocksRetrieval-augmented generation (RAG)Grounding a model's answers in your own documents and data by retrieving the relevant passages first and giving them to the model with the question.
S
- Building blocksSemantic layerA shared, governed definition of a company's business metrics and entities (revenue, customer, churn) that people, BI tools and AI can all query consistently.
- Strategy & adoptionShadow AIEmployees using AI tools that the organisation has not approved or does not know about, often by pasting company data into public chatbots.
- ModelsSmall language modelsAlso: SLMsLanguage models with far fewer parameters than frontier models, cheap and fast enough to run on modest hardware or on a device, often tuned for narrow tasks.
- Governance & regulationSovereign AIAI infrastructure, models and data under the jurisdiction and control of a country, region or organisation, rather than depending fully on foreign providers.
- Building blocksSynthetic dataArtificially generated data that mimics the patterns of real data, used to train, test or evaluate AI when real data is scarce, sensitive or unbalanced.
T
- Operations & costTokensThe small pieces of text (word fragments, words or symbols) that models read and write, and the unit in which most AI usage is priced.
- Agents & automationTool useAlso: function callingA model's ability to call external functions or APIs, such as a search, a database query or a payment step, and use the results in its answer.
U
V
W
No terms match.
Definitions are written by Scayl and cite primary sources where a claim needs one. Spotted something out of date? Tell us and we'll fix it.