AI Glossary

AI Glossary

Every term you'll encounter building, deploying, and scaling AI agents, defined clearly.

Filter:

Agent

A software system that perceives its environment, makes decisions, and takes actions autonomously to achieve a goal.

Related: AI Agent, Autonomous Agent

Agentic AI

AI systems designed to act autonomously with goal-directed behavior, planning, tool use, and iterative reasoning.

Related: AI Agent, Orchestration

AI Agent

An agent powered by artificial intelligence that can reason, plan, and execute tasks using language models and external tools.

Related: Agentic AI, Agent

API (Application Programming Interface)

A set of protocols and definitions that allow software components to communicate with each other programmatically.

Related: Integration Layer

Autonomous Agent

An agent that operates independently with minimal human oversight, making decisions and completing tasks end-to-end.

Related: Agent, Human-in-the-Loop

Baseline Model

A reference model used as a starting point for comparison when evaluating improvements from fine-tuning or new architectures.

Related: Foundation Model, Fine-Tuning

Batch Processing

Running AI inference or training tasks on collected groups of inputs rather than individually in real time.

Related: Inference, Workflow

Bot

A software application that automates repetitive tasks, often through predefined rules or AI-driven conversation.

Related: Chatbot, AI Agent

Chatbot

A conversational interface that interacts with users via text, often powered by NLP or large language models.

Related: Dialog Management, NLP

Confidence Score

A numerical value indicating how certain a model is about its prediction or output, typically between 0 and 1.

Related: Output Validation, Inference

Context Window

The maximum number of tokens a language model can process in a single interaction, including both input and output.

Related: Token, Tokenization

Conversation AI

AI systems designed to engage in natural, multi-turn dialogue with users, handling context and follow-up questions.

Related: Dialog Management, Multi-Turn Conversation

Deep Learning

A subset of machine learning using multi-layered neural networks to learn hierarchical representations from raw data.

Related: Neural Network, Transformer

Decision Tree

A supervised learning model that splits data into branches based on feature values to make predictions.

Related: Model, Training Data

Dialog Management

The component of a conversational system that tracks dialogue state and determines the next action or response.

Related: Intent Recognition, Multi-Turn Conversation

Embedding

A dense vector representation of text, images, or data in a continuous vector space, capturing semantic meaning.

Related: Vector Database, RAG

End-to-End Automation

Automating an entire workflow from initial input to final output with minimal or no human intervention.

Related: Workflow, Workforce Automation

Entity Recognition

Identifying and classifying named entities (people, organizations, locations) within unstructured text.

Related: NLP, Intent Recognition

ETL (Extract, Transform, Load)

A data pipeline process that extracts data from sources, transforms it, and loads it into a target system.

Related: Workflow, Unstructured Data

Fine-Tuning

Adapting a pre-trained model to a specific task or domain by continuing training on a smaller, targeted dataset.

Related: Foundation Model, Training Data

Foundation Model

A large-scale model pre-trained on broad data that can be adapted to a wide range of downstream tasks.

Related: Large Language Model (LLM), Fine-Tuning

Few-Shot Learning

Training or prompting a model with only a small number of labeled examples to generalize to a new task.

Related: Zero-Shot Learning, Prompt Engineering

Governance

Policies, processes, and oversight frameworks that ensure AI systems are used responsibly and compliantly.

Related: Guardrails, RBAC

Grounding

Anchoring AI outputs to verified, real-world data sources to reduce hallucinations and improve factual accuracy.

Related: RAG, Hallucination

Guardrails

Safety mechanisms and constraints that prevent AI systems from producing harmful, biased, or off-topic outputs.

Related: Governance, Output Validation

Hallucination

When a model generates plausible-sounding but factually incorrect or fabricated information.

Related: Grounding, Confidence Score

Harness

An operational framework that wraps around an AI model to manage inputs, outputs, tool calls, and workflow orchestration.

Related: Workflow, Orchestration

Human-in-the-Loop (HITL)

A design pattern where a human reviews, validates, or approves AI decisions at critical points in a workflow.

Related: Governance, Output Validation

Inference

The process of using a trained model to generate predictions or outputs from new input data.

Related: Model, Latency

Intent Recognition

Identifying the user's goal or purpose from their input text to drive the appropriate system response.

Related: NLP, Entity Recognition

Integration Layer

Middleware that connects AI systems to external APIs, databases, and enterprise tools for data exchange.

Related: API, Workflow

Knowledge Base

A structured repository of information that AI systems query to provide accurate, domain-specific responses.

Related: Knowledge Graph, RAG

Knowledge Graph

A network of entities and their relationships, enabling AI to reason over structured domain knowledge.

Related: Knowledge Base, Embedding

Large Language Model (LLM)

A neural network with billions of parameters trained on massive text corpora to understand and generate human language.

Related: Foundation Model, Transformer

Latency

The time delay between sending a request to an AI model and receiving the response.

Related: Inference, Token

Learning Rate

A hyperparameter controlling how much model weights are updated during each training step.

Related: Training Data, Fine-Tuning

Model

A mathematical representation trained on data to perform tasks like prediction, classification, or generation.

Related: Foundation Model, Training Data

Multi-Turn Conversation

A dialogue exchange spanning multiple back-and-forth interactions, requiring the system to maintain context.

Related: Context Window, Dialog Management

Memory

The ability of an AI agent to retain and recall information across interactions within or across sessions.

Related: Context Window, Multi-Turn Conversation

NLP (Natural Language Processing)

A field of AI focused on enabling computers to understand, interpret, and generate human language.

Related: Intent Recognition, Sentiment Analysis

Neural Network

A computing architecture inspired by biological neurons, composed of layers of interconnected nodes that learn patterns from data.

Related: Deep Learning, Transformer

Orchestration

Coordinating multiple AI agents, tools, and workflows into a unified, sequential or parallel process.

Related: Workflow, Integration Layer

Output Validation

Checking AI-generated outputs for correctness, safety, format compliance, and adherence to business rules before delivery.

Related: Guardrails, Confidence Score

Prompt

The input text or instruction given to a language model to guide its generation or behavior.

Related: Prompt Engineering

Prompt Engineering

The practice of designing and refining prompts to elicit desired outputs from language models.

Related: Prompt, Few-Shot Learning

PII (Personally Identifiable Information)

Any data that can identify a specific individual, requiring strict handling and redaction in AI systems.

Related: Governance, Guardrails

Query

A request sent to a model or database to retrieve or generate specific information.

Related: Prompt, RAG

Quality Score

A metric evaluating the accuracy, relevance, and reliability of an AI system's outputs.

Related: Confidence Score, Output Validation

RAG (Retrieval-Augmented Generation)

A technique that combines a language model with external knowledge retrieval to ground responses in verified data.

Related: Knowledge Base, Grounding

RBAC (Role-Based Access Control)

A security model that restricts system access based on user roles and permissions.

Related: Governance

Reinforcement Learning

A training paradigm where an agent learns optimal behavior by receiving rewards or penalties for its actions.

Related: Agent, Fine-Tuning

Sentiment Analysis

Using NLP to determine the emotional tone (positive, negative, neutral) of text input.

Related: NLP, Intent Recognition

Simulation

Running AI agents in a controlled virtual environment to test behavior before real-world deployment.

Related: Agent, Validation

Step

A single discrete action within a workflow, such as an API call, data transform, or model inference.

Related: Workflow, Orchestration

Structured Output

AI responses returned in a predefined schema (JSON, XML) for easy parsing and downstream processing.

Related: Output Validation, Query

Token

The smallest unit of text processed by a language model, which may be a word, subword, or character.

Related: Tokenization, Context Window

Tokenization

The process of breaking text into tokens that a language model can process.

Related: Token, NLP

Training Data

The dataset used to teach a model patterns and relationships during the training process.

Related: Model, Fine-Tuning

Transformer

A neural network architecture using self-attention mechanisms that powers most modern language models.

Related: Large Language Model (LLM), Deep Learning

Unstructured Data

Data without a predefined format, such as emails, documents, images, or audio files.

Related: Embedding, ETL

Vector Database

A specialized database optimized for storing and querying high-dimensional vector embeddings for similarity search.

Related: Embedding, RAG

Validation

Testing an AI model's performance on unseen data to assess generalization and detect overfitting.

Related: Output Validation, Quality Score

Workflow

A defined sequence of steps and decisions that an AI agent or system follows to complete a task.

Related: Orchestration, Step

Workforce Automation

Replacing or augmenting repetitive human tasks with AI agents to improve speed, consistency, and scale.

Related: Agent, End-to-End Automation

Zero-Shot Learning

A model's ability to perform a task it was never explicitly trained on, using only its general knowledge.

Related: Few-Shot Learning, Foundation Model