🧠 The Practical AI Glossary for Teams

🧠 Foundations

Artificial Intelligence (AI) — A broad field of computer science focused on building systems that can perform tasks typically requiring human intelligence — like recognizing speech, making decisions, or translating languages.

Machine Learning (ML) — A subset of AI where systems improve their performance by learning patterns from data, rather than being explicitly programmed with rules for every scenario.

Deep Learning — A type of machine learning that uses neural networks with many layers to tackle complex tasks like image recognition and language understanding. The "deep" refers to the depth of those layers.

Neural Network — A computational architecture loosely inspired by the human brain, made up of interconnected nodes that process and pass information to recognize patterns in data.

Natural Language Processing (NLP) — The field of AI concerned with enabling computers to understand, interpret, and generate human language — the foundation behind chatbots, translation, and text analysis.

Computer Vision — AI's ability to interpret and make sense of visual input — images, video, documents — the way a human would. Enables things like object detection and document scanning.

⚙️ Models and Generation

Model — The trained system itself — the result of feeding data through a learning algorithm. When you use ChatGPT or Claude, you're interacting with a model.

Large Language Model (LLM) — A model trained on vast amounts of text to understand and generate language. GPT-4, Claude, and Gemini are all LLMs. The "large" refers to the scale of both training data and model parameters.

Generative AI — AI systems capable of producing new content — text, images, audio, code, or video — rather than just analyzing or classifying existing content.

Fine-Tuning — Taking a pre-trained model and continuing its training on a smaller, task-specific dataset. This specializes the model for a particular domain or style without building from scratch.

Training Data — The dataset used to teach a model — the examples it learns patterns from. The quality and diversity of training data directly affects model behavior and blind spots.

Dataset — A structured collection of data — text, images, labels, or records — used to train, evaluate, or test an AI model.

Hallucination — When an AI model generates plausible-sounding but factually incorrect information — and presents it confidently. A known limitation of LLMs that requires verification for high-stakes outputs.

Temperature — A parameter that controls output randomness. Low temperature produces consistent, predictable responses; high temperature introduces more variety and creativity — at the cost of reliability.

🖥️ Prompting and Interaction

Prompt — The input you provide to an AI model — a question, instruction, or piece of context that triggers a response. Prompt quality strongly influences output quality.

System Prompt — Hidden instructions set by a developer before your conversation begins. They configure the AI's behavior, tone, and constraints without being visible to the end user.

Context Window — The total amount of text — conversation history, documents, and instructions — that an AI model can hold in working memory at once. Content outside this window is not accessible to the model.

Token — The smallest unit of text a model processes — roughly ¾ of a word. Models read and generate tokens, and context windows and pricing are both measured in them.

Retrieval-Augmented Generation (RAG) — A technique that grounds AI responses in retrieved documents or data sources rather than solely model memory — reducing hallucination and keeping answers factually current.

 🤖 Agents & Automation

AI Agent — An AI system that doesn't just respond to a single prompt, but takes a sequence of actions — using tools, browsing the web, writing code — to complete a multi-step goal.

Agentic AI — AI that operates with meaningful autonomy: it can plan, reason about sub-goals, use tools, and course-correct — rather than waiting for a human to direct each step.

Automation — Using AI — or software more broadly — to execute repetitive tasks with little or no human involvement. AI-driven automation handles tasks that previously required human judgment, not just rule-following.

Chatbot — A program that converses with users through natural language. Modern AI chatbots are powered by LLMs; older ones used rule-based scripts. The term covers a wide range of capability.

Voice AI — AI systems that process spoken language — combining speech recognition, language understanding, and text-to-speech to enable natural voice interaction.

Strategy and Ethics

Personalization — Using AI to tailor content, recommendations, or experiences to individual users based on their behavior, preferences, or history.

AI Ethics — The principles and practices governing how AI should be developed and used — covering fairness, accountability, transparency, privacy, and the prevention of harm.

Bias — Systematic errors in model outputs that reflect imbalances in training data or design choices — leading to unfair, skewed, or discriminatory results in some contexts.

AI Governance — The policies, processes, and oversight structures an organization puts in place to ensure AI systems are used safely, responsibly, and in alignment with legal and ethical standards.