AI Jargon Glossary
37+ AI terms explained in plain English. No technical background required. Designed for business owners and decision makers.
A
Agentic AI
AI systems that can independently plan, make decisions, and take actions to achieve goals, rather than just responding to single prompts. Think of it as AI that can manage a multi-step project rather than just answering a question. This is one of the fastest-growing areas in AI development.
Algorithm
A set of rules or instructions that a computer follows to solve a problem or complete a task. In AI, algorithms are the mathematical methods used to find patterns in data and make predictions. You do not need to understand algorithms to use AI tools.
API (Application Programming Interface)
A way for software applications to talk to each other. When a business integrates ChatGPT into their customer service system, they use OpenAI's API. Think of it as a waiter taking your order to the kitchen: you tell the API what you want, and it brings back the result.
B
Bias (AI)
When an AI system produces systematically unfair or prejudiced results. This usually happens because the training data contained existing biases. For example, if an AI is trained on historical hiring data that favoured certain demographics, it will reproduce those biases.
C
Chatbot
A software application that simulates conversation with users. Modern AI chatbots (like ChatGPT and Claude) use large language models to generate human-like responses, unlike older chatbots that followed rigid scripts.
Context Window
The maximum amount of text an AI model can consider at once, measured in tokens. Claude has a 200,000 token context window (roughly 150,000 words), meaning it can process entire books or codebases in one conversation. Larger context windows allow more complex tasks.
Computer Vision
AI that can interpret and understand images and video. Used for everything from quality control in manufacturing to analysing medical scans. In business, it powers things like automated document scanning and visual product inspection.
D
Data Labelling / Annotation
The process of tagging data with labels so AI systems can learn from it. For example, labelling photos of products as 'damaged' or 'undamaged' to train a quality control AI. This is often the most time-consuming part of building custom AI systems.
Deep Learning
A subset of machine learning that uses neural networks with many layers (hence 'deep') to learn from large amounts of data. It is the technology behind most modern AI breakthroughs, including language models and image generation.
DPIA (Data Protection Impact Assessment)
A formal assessment required under GDPR for data processing activities that pose high risks to individuals' privacy. If you use AI to process personal data at scale, you likely need to conduct one.
E
Embedding
A way of representing text, images, or other data as numbers (vectors) so that AI can understand relationships between them. Similar concepts end up close together in the number space. This is how AI 'understands' that 'king' and 'queen' are related concepts.
F
Few-Shot Learning
Teaching an AI to perform a task by giving it just a few examples, rather than training it on thousands. When you include 2-3 example outputs in your prompt, you are using few-shot learning. It is one of the most effective prompting techniques.
Fine-Tuning
The process of taking a pre-trained AI model and training it further on your specific data to make it better at a particular task. Like hiring a generalist and then training them in your specific industry. More expensive and complex than prompt engineering, but produces more consistent results.
Foundation Model
A large AI model trained on broad data that can be adapted for many different tasks. GPT-4, Claude, and Gemini are all foundation models. Think of them as general-purpose brains that can be directed toward specific tasks through prompting or fine-tuning.
G
Generative AI (GenAI)
AI that creates new content: text, images, code, audio, or video. ChatGPT, Claude, Midjourney, and DALL-E are all generative AI tools. This is the category of AI that has exploded in business use since 2023.
GPT (Generative Pre-trained Transformer)
The architecture behind OpenAI's models (GPT-4, GPT-4o). GPT is a specific type of large language model. The term is often used loosely to refer to any AI chatbot, but technically it refers specifically to OpenAI's products.
Grounding
Connecting AI outputs to verified, factual sources to reduce hallucinations. A grounded AI response cites where its information comes from, making it easier to verify. This is increasingly important for business applications.
Guardrails
Rules and restrictions built into AI systems to prevent harmful, inappropriate, or off-topic outputs. In business, guardrails might prevent an AI customer service bot from discussing competitors or making promises the company cannot keep.
H
Hallucination
When an AI generates information that sounds plausible but is factually incorrect. AI models predict likely text, not truthful text. They can confidently cite non-existent studies, invent statistics, or fabricate details. This is the single biggest risk for businesses using AI for factual content.
L
LLM (Large Language Model)
An AI model trained on vast amounts of text data that can understand and generate human language. ChatGPT, Claude, Gemini, and Llama are all LLMs. They are the engines behind most of the AI tools businesses use today.
M
Machine Learning (ML)
A type of AI where systems learn from data rather than being explicitly programmed. Instead of writing rules like 'if email contains these words, mark as spam,' you feed the system thousands of examples and it learns the patterns itself.
MCP (Model Context Protocol)
An open standard that allows AI models to connect to external tools, databases, and services. Think of it as a universal plug that lets AI assistants access your business systems directly, rather than you copy-pasting information back and forth.
Multimodal
AI models that can process and generate multiple types of content: text, images, audio, and video. GPT-4o and Gemini are multimodal, meaning you can show them an image and ask questions about it, or have them generate images alongside text.
N
Natural Language Processing (NLP)
The branch of AI focused on enabling computers to understand, interpret, and generate human language. Every time you type a question into ChatGPT, NLP is what allows the system to understand what you are asking.
Neural Network
A computing system inspired by the structure of the human brain, made up of interconnected nodes (neurons) that process information. Neural networks are the foundation of modern AI, enabling machines to recognise patterns in data.
O
Open Source AI
AI models whose code and weights are publicly available for anyone to use, modify, and deploy. Meta's Llama and Mistral are examples. Open source models give businesses more control over their data but require more technical expertise to deploy.
P
Prompt
The text instruction you give to an AI model. The quality of your prompt directly determines the quality of the output. Prompts can be simple questions or detailed instructions with context, examples, and constraints. Better prompts produce dramatically better results.
Prompt Engineering
The skill of crafting effective prompts to get the best possible output from AI models. It involves techniques like setting context, defining roles, providing examples, specifying formats, and adding constraints. It is the most immediately valuable AI skill for business professionals.
R
RAG (Retrieval-Augmented Generation)
A technique where AI retrieves relevant information from a specific knowledge base before generating a response. Instead of relying only on its training data, the AI first searches your documents, database, or help centre for relevant context. This dramatically reduces hallucinations for domain-specific questions.
Responsible AI
The practice of developing and using AI in ways that are ethical, fair, transparent, and accountable. This includes testing for bias, ensuring data privacy, maintaining human oversight, and being transparent about AI usage. Increasingly important for regulatory compliance.
S
Sentiment Analysis
AI that determines the emotional tone of text: positive, negative, or neutral. Used in business to analyse customer reviews, social media mentions, support tickets, and employee surveys at scale. Useful for monitoring brand perception.
T
Temperature
A setting that controls how creative or random an AI model's responses are. Low temperature (0.1 to 0.3) produces more predictable, focused outputs. High temperature (0.7 to 1.0) produces more varied, creative responses. For business writing, lower temperatures usually work better.
Token
The basic unit that AI models use to process text. A token is roughly three-quarters of a word in English. 'ChatGPT' might be two tokens. Pricing for AI APIs is often based on tokens processed. Understanding tokens helps you estimate costs and understand context window limits.
Training Data
The dataset used to teach an AI model. The quality, size, and diversity of training data directly affects the model's capabilities and biases. GPT-4 was trained on trillions of tokens from books, websites, and other text sources.
Transformer
The neural network architecture behind modern language models (the 'T' in GPT). Introduced by Google in 2017, it revolutionised AI's ability to understand context and relationships in text. You do not need to understand how transformers work to use AI effectively.
V
Vector Database
A specialised database designed to store and search embeddings (numerical representations of data). Used in RAG systems to quickly find relevant information from large document collections. If you are building custom AI applications, you will likely encounter vector databases.
Z
Zero-Shot Learning
When an AI performs a task it was not specifically trained for, just by understanding the instruction. If you ask ChatGPT to classify customer emails by topic without providing examples, that is zero-shot learning. It works because LLMs have broad general knowledge.
Frequently Asked Questions
Do I need to understand all these terms to use AI?
Absolutely not. You can get significant value from AI tools without understanding any of the technical terms. However, knowing the basics (like LLM, prompt, hallucination, and token) will help you make better decisions about which tools to use, understand limitations, and communicate more effectively with technical partners.
What is the most important AI term for business owners to know?
Hallucination. Understanding that AI can generate confident-sounding but incorrect information is the single most important thing for any business user. It changes how you approach quality control, fact-checking, and trust in AI outputs.
How quickly is AI terminology changing?
Rapidly. New terms appear monthly as the technology evolves. Terms like 'agentic AI,' 'MCP,' and 'vibe coding' barely existed a year ago. We update this glossary regularly, but the core concepts (LLM, NLP, tokens, prompts) are stable and worth understanding.
What is the difference between AI and machine learning?
AI is the broad goal of creating systems that can perform tasks requiring human intelligence. Machine learning is a specific approach to achieving AI by training systems on data rather than programming explicit rules. All machine learning is AI, but not all AI is machine learning. Think of it like: all squares are rectangles, but not all rectangles are squares.
Want to Go Beyond the Jargon?
Understanding the terms is a great start. If you want practical help implementing AI in your business, book a free discovery call.
Book a Free Call