Live Webinar On: Building AI-First Financial InstitutionsRegister Now →

AI Glossary

The AI Glossary for Enterprise Teams

150+ definitions covering agentic AI, LLMs, deployment, and enterprise AI infrastructure.

Featured Terms

Start here.

COREAgentic AI

Agentic AI

AI systems that reason, plan, and complete multi-step tasks autonomously.

Read definition

COREAgentic AI

AI Agent

A software system using a language model to understand a goal and take actions.

Read definition

COREEnterprise

Enterprise AI

AI systems built for the scale, security, and compliance of large regulated organizations.

Read definition

COREAgentic AI

Model Context Protocol (MCP)

The emerging standard for connecting AI agents to enterprise systems and data sources.

Read definition

COREDeployment

On-Premise AI

AI models deployed inside your own data center. Zero external data flow.

Read definition

COREFoundations

Large Language Model

A neural network trained on vast text corpora to understand and generate natural language.

Read definition

A20 terms

Foundations

Activation Function

A mathematical function that determines how strongly a neuron responds and gives neural networks the non-linearity needed to learn complex patterns.

Read

Data & Training

Active Learning

A machine learning approach in which the model requests labels for the most informative or uncertain examples first.

Read

Data & Training

Adversarial Training

A robustness technique that trains a model on intentionally perturbed or malicious examples so it performs better against attacks and edge cases.

Read

Agentic AI

AI systems that reason, plan, and complete multi-step tasks autonomously.

Read

Agentic AI

AI Agent

A software system using a language model to understand a goal and take actions.

Read

Agentic AI

AI Assistant

A conversational AI system that helps users complete tasks through natural language.

Read

B4 terms

Foundations

Backpropagation

The core algorithm used to train modern neural networks and large language models.

Read

Data & Training

Batch Size

The number of training examples processed before a model updates its parameters once.

Read

Foundations

Bayesian Networks

Probabilistic graphical models that represent variables and their conditional dependencies as a directed acyclic graph.

Read

NLP & Language

BERT

A foundational bidirectional language model architecture from Google, released in 2018.

Read

C11 terms

NLP & Language

Chain of Thought Prompting

A technique for improving LLM reasoning by breaking complex problems into intermediate steps.

Read

Agentic AI

Chatbot

A rule-based text conversation system, distinct from agentic AI which reasons and acts.

Read

Data & Training

Chunking

The process of dividing long documents or data streams into smaller units that models can retrieve and process effectively.

Read

Foundations

CNN (Convolutional Neural Network)

A neural network architecture that uses learned filters to recognise spatial patterns in images, video, audio, and other grid-like data.

Read

Agentic AI

Cognitive Architecture

A structured design for combining perception, memory, reasoning, planning, and action in an intelligent system.

Read

Data & Training

Collaborative Filtering

A recommendation technique that predicts preferences from similarities in how users interact with items.

Read

D14 terms

Data & Training

Data Annotation

Labelling raw data with tags so machine learning models can learn from it.

Read

Data & Training

Data Augmentation

The creation of realistic variations of training examples to improve model generalisation without collecting equivalent new data.

Read

Enterprise

Data Lake

A scalable repository that stores large volumes of raw structured, semi-structured, and unstructured data in its native format.

Read

Data & Training

Data Preprocessing

The cleaning, transformation, validation, and preparation of raw data before it is used for analytics or model training.

Read

Enterprise

Data Privacy

The principles and controls governing how personal or sensitive data is collected, used, shared, retained, and deleted.

Read

Enterprise

Data Security

The technical and organisational safeguards that protect data confidentiality, integrity, and availability throughout its lifecycle.

Read

E7 terms

Deployment

Edge AI

Deployment of AI models directly on local devices instead of centralized cloud.

Read

Foundations

Embedding

A numerical representation of text or data that captures semantic meaning.

Read

Data & Training

Ensemble Learning

A machine learning strategy that combines several models to produce more accurate or robust predictions than a single model.

Read

Enterprise

Enterprise AI

AI systems built for the scale, security, and compliance of large regulated organizations.

Read

Data & Training

Epochs

Complete passes through a training dataset during which a model has an opportunity to learn from every example.

Read

Enterprise

ETL (Extract Transform Load)

A data integration process that extracts data from source systems, transforms it into a usable form, and loads it into a target repository.

Read

F8 terms

Data & Training

F1 Score

A classification metric that balances precision and recall using their harmonic mean.

Read

Data & Training

Feature Engineering

The creation, transformation, and selection of input variables that make useful patterns easier for a model to learn.

Read

Data & Training

Feature Selection

The process of retaining the most informative input variables while removing irrelevant or redundant ones.

Read

Deployment

Federated Learning

A distributed training approach that learns from data across multiple locations without centralising the underlying records.

Read

Data & Training

Few-Shot Learning

A model's ability to perform or learn a task from only a small number of labelled examples or demonstrations.

Read

Data & Training

Fine-Tuning

Adapting a pre-trained model on domain-specific data to improve task performance.

Read

G7 terms

Foundations

Generative Adversarial Network (GAN)

A generative architecture in which a generator and discriminator compete to create increasingly realistic synthetic data.

Read

Foundations

Generative AI

AI that creates new content — text, images, code, or audio — from learned patterns.

Read

Foundations

Gradient Descent

An optimisation algorithm that iteratively changes model parameters in the direction that reduces prediction error.

Read

Foundations

Graph RAG

A retrieval-augmented generation approach that uses relationships in a knowledge graph to retrieve connected evidence for an answer.

Read

Data & Training

Ground Truth

The trusted reference labels or measurements used to train a model and evaluate whether its outputs are correct.

Read

Foundations

Grounding

The practice of connecting an AI response to verified data, retrieved evidence, or real-world context rather than relying only on model memory.

Read

H4 terms

Foundations

Hallucination

When an AI model generates confident but factually incorrect or fabricated output.

Read

Data & Training

Hierarchical Clustering

An unsupervised method that builds a tree of nested groups by progressively merging or splitting similar observations.

Read

Enterprise

Human-in-the-Loop

A design pattern where AI decisions above a risk threshold require human review.

Read

Data & Training

Hyperparameter Tuning

The systematic search for model configuration values that produce the best validated performance.

Read

I4 terms

Foundations

In-Context Learning

Teaching an LLM a new task by showing examples in the prompt, without retraining.

Read

Deployment

Inference

Running a trained AI model to generate predictions or outputs from new inputs.

Read

NLP & Language

Information Retrieval

The methods used to index, search, rank, and return information that is relevant to a query.

Read

Deployment

Infrastructure as a Service (IaaS)

Cloud computing infrastructure delivered on demand, including virtual machines, storage, networking, and accelerators.

Read

J1 term

Foundations

JSON Mode

A model output format that guarantees structured JSON, enabling reliable downstream parsing.

Read

K4 terms

Data & Training

K-Means Clustering

An unsupervised algorithm that assigns observations to a chosen number of clusters based on distance from each cluster centre.

Read

Enterprise

Knowledge Base

A structured repository of enterprise content used to ground AI agent responses.

Read

Data & Training

Knowledge Distillation

A model compression technique that trains a smaller student model to reproduce the behaviour of a larger teacher model.

Read

Foundations

Knowledge Graph

A structured network of entities, facts, and relationships that gives machines an explicit representation of connected knowledge.

Read

L6 terms

Foundations

Large Language Model

A neural network trained on vast text corpora to understand and generate natural language.

Read

Deployment

Latency

The time delay between an AI input and its response. Critical for real-time applications.

Read

Data & Training

Learning Rate

The hyperparameter that controls how much a model changes its parameters after each optimisation step.

Read

Deployment

LLMOps

The operational practices for evaluating, deploying, monitoring, securing, and improving large language model applications.

Read

Data & Training

LoRA (Low-Rank Adaptation)

A parameter-efficient fine-tuning method that adapts a large model by training small low-rank matrices instead of all model weights.

Read

Foundations

Loss Function

A mathematical measure of prediction error that gives model training a quantity to minimise.

Read

M12 terms

Foundations

Machine Learning

A field of AI in which systems learn patterns from data to make predictions, decisions, or generate outputs without explicit rules for every case.

Read

NLP & Language

Machine Translation

The automatic translation of text or speech from one natural language into another.

Read

Agentic AI

Memory (in AI)

The mechanisms that let an AI system retain relevant information across turns, tasks, sessions, or long-running workflows.

Read

Data & Training

Meta Learning

An approach that trains models to adapt to new tasks quickly by learning reusable strategies for learning.

Read

Deployment

MLOps

The engineering discipline for versioning, deploying, monitoring, governing, and continuously improving machine learning systems.

Read

Agentic AI

Model Context Protocol (MCP)

The emerging standard for connecting AI agents to enterprise systems and data sources.

Read

N5 terms

NLP & Language

Named Entity Recognition

An NLP task that identifies and classifies named items such as people, organisations, locations, products, and dates in text.

Read

NLP & Language

Natural Language Generation

The use of AI to produce human-readable text from data, instructions, retrieved evidence, or internal representations.

Read

NLP & Language

Natural Language Processing

The AI field covering machines' ability to understand, process, and generate human language.

Read

Foundations

Neural Network

A computational model loosely inspired by the brain, used to learn patterns from data.

Read

NLP & Language

NLU (Natural Language Understanding)

The set of techniques that helps software interpret the meaning, intent, entities, and context in human language.

Read

O3 terms

Deployment

On-Premise AI

AI models deployed inside your own data center. Zero external data flow.

Read

Agentic AI

Orchestration

Coordinating multiple AI agents and tools to complete complex multi-step enterprise workflows.

Read

Data & Training

Overfitting

A failure mode in which a model learns training examples too closely and performs poorly on new data.

Read

P8 terms

NLP & Language

Parsing

The analysis of text or structured input to identify its components, syntax, and relationships in a machine-usable form.

Read

Deployment

Platform as a Service (PaaS)

A managed cloud environment for building and running applications without operating the underlying servers and runtime infrastructure.

Read

Data & Training

Precision and Recall

Complementary classification metrics that measure how many positive predictions are correct and how many actual positives are found.

Read

Enterprise

Predictive Analytics

The use of historical and current data to estimate likely future outcomes, risks, behaviours, or demand.

Read

Enterprise

Prescriptive Analytics

Analytics that recommends which action to take by combining predictions, constraints, objectives, and business rules.

Read

Data & Training

Principal Component Analysis (PCA)

A dimensionality-reduction technique that transforms correlated variables into a smaller set of orthogonal components.

Read

Q1 term

Deployment

Quantization

Reducing model weight precision to shrink size and accelerate inference on constrained hardware.

Read

R9 terms

Foundations

RAG (Retrieval-Augmented Generation)

Retrieval Augmented Generation. Grounding LLM outputs in retrieved enterprise documents.

Read

Data & Training

Random Forest

An ensemble of decision trees whose combined predictions improve accuracy and reduce sensitivity to individual training samples.

Read

Agentic AI

Reasoning Model

A model optimised to work through multi-step problems, evaluate alternatives, and produce or execute a justified conclusion.

Read

Data & Training

Regularization

Techniques that discourage unnecessary model complexity so a model generalises better to unseen data.

Read

Data & Training

Reinforcement Learning

A learning approach in which an agent improves its behaviour through rewards and penalties received from an environment.

Read

Enterprise

Responsible AI

The practice of designing, deploying, and governing AI so it is fair, safe, transparent, privacy-preserving, and accountable.

Read

S9 terms

NLP & Language

Semantic Search

Search that matches meaning rather than keywords, powered by embeddings.

Read

NLP & Language

Sentiment Analysis

An NLP task that identifies the opinion, emotion, or attitude expressed in text or speech.

Read

Deployment

Small Language Model (SLM)

A compact language model designed for efficient, specialised, private, or edge deployment with lower compute requirements.

Read

Deployment

Sovereign AI

AI infrastructure that stays inside a country or organization's perimeter.

Read

NLP & Language

Speech Recognition

The automatic conversion of spoken language into text or structured commands.

Read

Foundations

Supervised Learning

Machine learning in which a model learns a mapping from inputs to known target labels or values.

Read

T7 terms

Agentic AI

Task Decomposition

The process of breaking a complex goal into smaller, ordered, and verifiable steps that an agent can execute.

Read

NLP & Language

Text Classification

The automatic assignment of categories, intents, risks, or labels to text.

Read

Foundations

Token

The basic unit of text that language models process. Roughly 3/4 of an English word.

Read

NLP & Language

Tokenization

The conversion of text or other input into the discrete units a model processes.

Read

Agentic AI

Tool Use

The ability of an AI agent to call external APIs, run code, or query databases.

Read

Foundations

Transfer Learning

The reuse of knowledge learned on one task or dataset to improve performance on another.

Read

U2 terms

Data & Training

Underfitting

A failure mode in which a model is too simple or insufficiently trained to capture important patterns in the data.

Read

Foundations

Unsupervised Learning

Machine learning that discovers patterns, groups, or representations in data without labelled target answers.

Read

V2 terms

Foundations

Vector Database

A database optimised for storing and searching embeddings, enabling semantic retrieval.

Read

Agentic AI

Voice AI

Real-time conversational AI over voice channels. Multilingual, natural.

Read

W1 term

Enterprise

Workflow Automation

Using AI to automate multi-step business processes across systems and teams.

Read

Z1 term

Foundations

Zero-Shot Learning

A model's ability to handle tasks or categories it has never seen during training.

Read

Beyond the Glossary

Understanding the terms is one thing.

Putting enterprise AI into production is another. See how Fluid AI runs agentic AI at scale.

Explore the Platform Book a Demo