AI glossary
Plain-English definitions of the artificial intelligence and automation terms that come up when a business evaluates, builds, and runs AI. From foundations like machine learning to production concepts like retrieval, agents, and orchestration.
Search a term or jump to a letter. Each entry links to a fuller explainer.
119 terms
A
- Agentic AI Also: autonomous AI, AI agents Agents & Orchestration AI that pursues a goal across multiple steps on its own, choosing what to do next and acting through connected tools instead of waiting for a prompt each turn.
- Agentic Workflow Also: agent workflow Agents & Orchestration A process where an AI agent decides its own steps to reach a goal, rather than following a fixed sequence set in advance.
- AI Agent Also: autonomous agent, agent Agents & Orchestration A software system that pursues a goal on its own, deciding what to do next and taking action through tools rather than only producing text.
- AI Automation Also: intelligent automation Automation & Process The use of AI to handle work that fixed rules cannot, so a process can run with less human involvement.
- AI Consulting Also: AI advisory Business & Engagement Professional advice that helps an organization decide where and how to use AI, starting from the business problem and the return rather than a specific tool.
- AI Copilot Also: copilot, AI assistant Agents & Orchestration An assistant that works alongside a person inside their existing tools, suggesting and drafting while the person stays in control.
- AI Governance Also: AI oversight Governance & Risk The policies, controls, and accountability that decide how an organization builds, deploys, and monitors AI responsibly.
- AI Integration Also: system integration Deployment & Ops The work of connecting an AI model to the systems, data, and tools a business already uses, so it can act on real information rather than run in isolation.
- AI Maturity Model Also: AI adoption model Business & Engagement A framework describing the stages an organization moves through as it adopts AI, from early experiments to AI embedded in daily operations.
- AI Readiness Assessment Also: AI readiness Business & Engagement A structured review of whether an organization has the data, processes, systems, and use cases it needs to deploy AI successfully.
- API Also: application programming interface Deployment & Ops A set of rules that lets two software systems talk to each other, defining how one asks another for data or an action and what it gets back.
- Artificial General Intelligence Also: AGI, strong AI Foundations A hypothetical AI that could understand and learn any intellectual task a human can, rather than being limited to specific ones. It does not exist today.
- Attention Mechanism Also: self-attention Models & Architecture The part of a modern AI model that lets it weigh which pieces of the input matter most, the core idea behind the transformer.
B
- Batch Inference Also: batch processing Deployment & Ops Running an AI model over a large set of inputs at once, on a schedule or in bulk, rather than responding to each request the moment it arrives.
- Benchmark Also: model benchmark, eval benchmark Models & Architecture A standardized test used to measure and compare how well AI models perform on a specific task, using a fixed dataset and scoring method.
- Bias Also: algorithmic bias, model bias Governance & Risk A systematic skew in a model's outputs that unfairly favors or disadvantages certain groups or outcomes, usually inherited from training data.
- Business Process Automation Also: BPA Automation & Process The use of software to run an entire business process from start to finish with little or no manual work.
C
- Chain-of-Thought Prompting Also: CoT Prompting A technique that asks an AI model to work through a problem step by step before giving its final answer, improving accuracy on multi-step tasks.
- Chatbot Also: conversational bot Agents & Orchestration A software program that holds a conversation with a person through text or voice, answering questions in natural language.
- Chunking Also: text chunking Retrieval & Data Splitting a large document into smaller pieces so an AI system can search and retrieve the most relevant parts, a key step in RAG.
- Closed-Source Model Also: proprietary model, closed-weight model Models & Architecture An AI model whose internal weights and code are kept private and used only through a provider's API.
- Computer Vision Also: CV Models & Architecture A field of AI that lets software interpret images and video, identifying objects, text, patterns, and activity.
- Context Engineering Also: context design Prompting The practice of deciding what information an AI model receives at runtime, so it has exactly what it needs to give a good answer.
- Context Window Also: context length Models & Architecture The amount of text an AI model can consider at one time, measured in tokens, covering the prompt, data, and conversation so far.
- Conversational AI Also: conversational interface Agents & Orchestration Technology that lets software understand and respond to human language in a natural, back-and-forth way, across text or voice.
- Custom AI Also: bespoke AI Business & Engagement An AI system built for a specific organization and its workflows, designed around its own data, processes, and goals, and typically owned by that company.
D
- Data Annotation Also: data labeling, labeling Retrieval & Data Labeling raw data, such as images, text, audio, or video, so a machine learning model can learn from it.
- Data Governance Governance & Risk The policies, roles, and processes an organization uses to manage its data's availability, quality, security, and compliance.
- Data Pipeline Retrieval & Data An automated series of steps that moves data from source to destination, cleaning and reshaping it along the way.
- Data Readiness Retrieval & Data A measure of how prepared an organization's data is to support an AI project: accessible, accurate, organized, and permitted for use.
- Deep Learning Also: DL Foundations A type of machine learning that uses many-layered neural networks to find patterns in data, learning features on its own.
- Deployment Also: model deployment, production deployment Deployment & Ops Putting a trained AI model into live use, where it responds to real requests from users or other systems.
- Distillation Also: model distillation, knowledge distillation Models & Architecture Compressing a large AI model into a smaller one by training the smaller model to copy the larger one's outputs.
E
- Edge Case Deployment & Ops A rare or extreme input that falls outside what a model was trained on, causing it to behave unpredictably or fail.
- Embedding Also: vector embedding, embeddings Retrieval & Data A list of numbers that represents the meaning of data such as a word, sentence, or image, so software can compare items by similarity.
- Embedding Model Retrieval & Data An AI model that turns data such as text or images into embeddings, the numeric vectors that capture meaning.
- Evaluation Also: evals, model evaluation Deployment & Ops Measuring how well an AI model performs, using tests and metrics to judge its accuracy, reliability, and fitness for a task.
- Explainability Also: interpretability Governance & Risk The degree to which a person can understand why an AI model produced a given output.
F
- Few-Shot Prompting Also: few-shot learning Prompting Guiding an AI model by including a small number of worked examples in the prompt, so it follows the pattern for a new input.
- Fine-Tuning Models & Architecture Taking a pretrained AI model and training it further on a smaller, specific dataset so it performs better on a particular task or domain.
- Foundation Model Models & Architecture A large AI model trained on broad data that serves as a general-purpose base, then adapted to many specific tasks.
- Frontier Model Models & Architecture One of the most advanced AI models available at a given time, at the leading edge of scale and capability.
- Function Calling Agents & Orchestration A capability that lets an AI model trigger external tools by producing a structured request the surrounding software runs.
G
- Generative AI Also: GenAI Foundations AI that creates new content such as text, images, audio, or code, rather than only analyzing or classifying existing data.
- GPU Also: graphics processing unit Deployment & Ops A specialized chip that performs many calculations at once, the hardware behind most AI training and much of running models.
- Grounding Also: AI grounding Retrieval & Data Connecting an AI model's responses to verifiable external information so answers are based on real data, not just training patterns.
- Guardrails Also: AI guardrails Governance & Risk The controls and rules that keep an AI system's behavior safe, appropriate, and within intended limits.
H
- Hallucination Also: AI hallucination, confabulation Governance & Risk When an AI model produces false or fabricated information presented as if it were true.
- Human-in-the-Loop Also: HITL Governance & Risk An approach where people review, approve, or correct an AI system's output at key points rather than letting it run fully on its own.
- Hybrid Search Retrieval & Data A retrieval method that combines keyword search with semantic search to return better results than either alone.
- Hyperparameter Models & Architecture A setting chosen before training that controls how an AI model learns, such as the learning rate or number of training passes.
I
- Inference Also: model inference Deployment & Ops The stage where a trained AI model is used to produce outputs on new data — the "using" phase, as opposed to training.
- Intelligent Automation Also: IA Automation & Process The use of AI together with automation tools to handle complex, judgment-based processes from end to end.
- Intelligent Document Processing Also: IDP Automation & Process Using AI to extract, classify, and interpret information from documents such as invoices, forms, and contracts.
- Intent Recognition Also: intent classification Models & Architecture Identifying what a user wants from their natural-language input, mapping many phrasings to a single defined goal.
J
K
- Knowledge Base Also: KB Retrieval & Data An organized store of trusted information an AI system retrieves from to ground its answers.
- Knowledge Cutoff Also: training cutoff Models & Architecture The date after which an AI model has no built-in information, because its training data ended there.
- Knowledge Graph Retrieval & Data A structured network that represents information as entities and the relationships between them.
L
- Large Language Model Also: LLM Foundations An AI model trained on huge amounts of text to understand and generate human language.
- Latency Deployment & Ops The delay between making a request and getting a response — for AI, the time from prompt to answer.
- LLMOps Deployment & Ops The practices and tools for building, deploying, and maintaining large language model applications in production.
- Long Context Also: extended context Models & Architecture An AI model's ability to take in and work with a very large amount of input at once, enabled by a large context window.
M
- Machine Learning Also: ML Foundations A branch of AI in which systems learn patterns from data rather than following rules written by hand.
- Memory Also: agent memory, conversational memory Agents & Orchestration An AI system's ability to retain and reuse information across turns or sessions, beyond a single prompt.
- MLOps Deployment & Ops The practices for deploying, monitoring, and maintaining machine learning models in production reliably.
- Model Context Protocol Also: MCP Agents & Orchestration An open standard for connecting AI models to external tools and data sources through a single, consistent interface.
- Model Drift Also: drift, concept drift Deployment & Ops The gradual decline in an AI model's performance over time as real-world data moves away from its training data.
- Multi-Agent System Also: MAS Agents & Orchestration A setup where several AI agents work together, each handling part of a task and coordinating to reach a shared goal.
- Multimodal AI Models & Architecture AI that can work with more than one type of data — text, images, audio, and video — together.
N
- Named Entity Recognition Also: NER Models & Architecture An NLP task that finds and classifies named things in text, such as people, organizations, places, dates, and amounts.
- Natural Language Processing Also: NLP Foundations The field of AI focused on getting computers to understand, interpret, and generate human language.
- Neural Network Foundations A computing system loosely inspired by the brain, made of connected layers of nodes that learn patterns from data.
- No-Code / Low-Code Also: low-code Business & Engagement Approaches to building software using visual tools instead of writing code by hand.
O
- Observability Also: AI observability Deployment & Ops The ability to see into how an AI system behaves in production, by collecting data on its inputs, outputs, performance, cost, and errors.
- Ontology Retrieval & Data A formal, structured definition of the concepts in a domain and the relationships between them.
- Open-Source Model Also: open-weight model Models & Architecture An AI model whose weights, and often its code, are publicly available to download, run, modify, and build on.
- Optical Character Recognition Also: OCR Automation & Process Technology that converts images of text, such as scans, photos, and PDFs, into machine-readable text.
- Orchestration Also: AI orchestration Agents & Orchestration The coordination of the parts of an AI system — models, tools, data, and steps — into a working process that produces a result.
- Overfitting Models & Architecture When a model learns its training data too closely, memorizing noise instead of general patterns, so it does well on training data but poorly on new data.
P
- Parameter Also: model parameter, weights Models & Architecture One of the internal values a model learns during training, also called a weight; a model's parameters store what it learned.
- Pre-Training Models & Architecture The initial, large-scale training of an AI model on broad data to learn general patterns, before it is adapted to any specific task.
- Predictive AI Also: predictive analytics Foundations AI that forecasts outcomes or classifies data based on patterns in past information, rather than creating new content.
- Process Mining Automation & Process A technique that analyzes the event logs recorded by business software to discover how a process actually runs, step by step.
- Prompt Prompting The input you give an AI model to get a response, usually text containing a question, instruction, or example.
- Prompt Chaining Prompting Breaking a complex task into a sequence of prompts, where the output of one becomes the input to the next.
- Prompt Engineering Prompting The practice of writing and refining the prompts given to an AI model to get more accurate, useful, and consistent output.
- Prompt Injection Governance & Risk An attack where hidden instructions in content an AI model reads trick it into ignoring its task and following the attacker's commands.
Q
- Quantization Models & Architecture A technique that shrinks an AI model by storing its numbers at lower precision, making it smaller and faster with a small loss in accuracy.
- Question Answering Also: QA Models & Architecture An AI task where a system responds to natural-language questions with direct answers, rather than a list of links.
R
- Rate Limit Deployment & Ops A rate limit is a cap on how many requests you can make to a service or API within a set period, such as a number of calls per minute. It protects the service from overload and shares capacity fairly across everyone using it.
- Reasoning Model Also: thinking model Models & Architecture A reasoning model is a large language model designed to work through a problem step by step before answering, spending extra computation on internal reasoning to handle harder, multi-step tasks more reliably.
- Reinforcement Learning Also: RL Foundations Reinforcement learning (RL) is a type of machine learning where a system learns by trial and error. It takes actions in an environment, receives rewards or penalties depending on the outcome, and adjusts its behavior to earn more reward over time.
- Reinforcement Learning from Human Feedback Also: RLHF Models & Architecture Reinforcement learning from human feedback (RLHF) is a technique for training AI models using people's preferences. Humans rate or compare model outputs, and those judgments are used to steer the model toward responses people find more helpful, honest, and safe.
- Reranking Retrieval & Data Reranking is a second step in a search or retrieval pipeline that reorders an initial set of results by relevance, using a more precise model to promote the best matches to the top.
- Retrieval-Augmented Generation Also: RAG Retrieval & Data Retrieval-augmented generation (RAG) is a technique that retrieves relevant information from a data source and gives it to a language model as context, so the model answers from real, current data rather than memory alone.
- Robotic Process Automation Also: RPA Automation & Process Robotic process automation (RPA) uses software bots to carry out repetitive, rule-based digital tasks the way a person would, such as moving data between systems, by following a fixed script.
S
- Semantic Search Retrieval & Data Semantic search finds results by meaning rather than by matching exact keywords. It represents the query and the content as embeddings and returns the items whose meaning is closest to the query.
- Sentiment Analysis Retrieval & Data Sentiment analysis is a natural-language processing task that determines the emotional tone of a piece of text, classifying it as positive, negative, or neutral, so a system can gauge how people feel at scale.
- Service Level Agreement Also: SLA Business & Engagement A service level agreement (SLA) is a contract that defines the level of service a provider commits to deliver, such as uptime, response time, and support availability, with measurable targets and remedies if they are missed.
- Small Language Model Also: SLM Models & Architecture A small language model (SLM) is a compact language model with far fewer parameters than a large one, making it cheaper to run, faster to respond, and able to work on modest hardware while still handling many focused tasks well.
- Structured Output Agents & Orchestration Structured output is when an AI model returns its response in a defined, machine-readable format, such as JSON that follows a set schema, rather than as free-form text, so other software can read specific fields and act on them.
- Supervised Learning Foundations Supervised learning is a type of machine learning that trains on labeled examples, where each input is paired with the correct answer, then uses the learned relationship to predict the answer for new data it has not seen.
- Synthetic Data Retrieval & Data Synthetic data is artificially generated data that imitates the patterns and structure of real data, used to train or test AI models when real data is scarce, sensitive, or costly to gather.
- System Prompt Prompting A system prompt is a set of background instructions that define an AI model's role, tone, and rules, applied before any user request, shaping how the model behaves across the whole conversation.
T
- Temperature Models & Architecture Temperature is a setting that controls how random or predictable a language model's output is. A lower temperature gives focused, consistent responses; a higher temperature gives more varied and creative output.
- Token Models & Architecture A token is the basic unit of text a language model processes, often a whole word or a piece of one. Models generate output one token at a time, and token counts determine how much fits in a model's context and what a request costs.
- Tokenization Models & Architecture Tokenization is the process of breaking text into tokens, the small units a language model reads, such as whole words or pieces of words. It is the first step in turning raw text into something a model can process.
- Tool Use Agents & Orchestration Tool use is an AI model's ability to call external tools, such as a search engine, a calculator, an API, or a database, to fetch information or take actions beyond generating text.
- Training Data Retrieval & Data Training data is the collection of examples an AI model learns from during training. Its quality, quantity, and coverage largely determine how well the model performs and where it falls short.
- Transformer Model Models & Architecture A transformer model is a type of neural network architecture that uses a mechanism called attention to weigh the relationships between all parts of its input at once. It is the architecture behind most modern large language models.
U
- Unstructured Data Retrieval & Data Unstructured data is information that has no predefined format or consistent organization, such as emails, documents, images, audio, and video. It makes up most of the data organizations hold but does not fit a fixed schema.
- Unsupervised Learning Foundations Unsupervised learning is a type of machine learning that finds patterns and structure in unlabeled data, without being told the correct answers. It discovers how the data organizes itself rather than predicting a known target.
V
- Vector Retrieval & Data A vector is a list of numbers that represents data as a point in a multi-dimensional space. In AI, vectors are how information is encoded numerically, and the distance between two vectors measures how similar the underlying data is.
- Vector Database Retrieval & Data A vector database is a database built to store embeddings and search them by similarity, quickly finding the items whose meaning is closest to a query. It is the storage and retrieval layer behind semantic search and RAG.
- Vertical AI Business & Engagement Vertical AI is artificial intelligence built for a specific industry or domain, shaped around its particular data, terminology, workflows, and rules, in contrast to general-purpose horizontal AI.
- Voice AI Agents & Orchestration Voice AI is artificial intelligence that understands and generates spoken language, letting people interact with systems by talking rather than typing. It powers voice assistants and automated phone systems.
W
Z
Frequently asked questions.
What the glossary is, who it's for, and how it stays current.
What is an AI glossary?
An AI glossary is a reference that defines artificial intelligence and automation terms in plain language. This one covers the terms a business runs into while evaluating, buying, or building AI, each with a short definition and a link to a fuller explainer.
Who is this glossary for?
Business decision-makers, not researchers. Definitions favor practical meaning and clarity over mathematical detail.
What is the difference between AI and machine learning?
Artificial intelligence is the broad goal of software that performs tasks needing human-like intelligence. Machine learning is one approach to it, where systems learn patterns from data instead of being explicitly programmed.
How often does the glossary change?
Terms are added and revised as the field moves. It is a living page: every entry carries its own definition and links to related terms, so the set stays connected as it grows.
Want to put AI
to work?
We work with leadership teams to find the right opportunities, define the strategy, and build the systems that move the business forward.
