AI jargon, without the hype
AI is moving fast and the vocabulary is half the battle. Here's what the common terms actually mean for a business, in plain English. Search for anything.
LLM
The type of AI behind tools like Claude and ChatGPT. It predicts text, which lets it write, summarise and answer.
Token
A chunk of text (roughly ¾ of a word) that models read and write in. Pricing and limits are measured in tokens.
Prompt
The instruction or question you give an AI. Better prompts get better results.
Prompt engineering
The craft of writing prompts, and giving examples and context, to get reliable output.
RAG
Giving an AI access to your own documents so its answers are grounded in your facts, not just its training.
Hallucination
When an AI states something confidently that's simply wrong. Why human review still matters.
Fine-tuning
Further-training a model on your specific data so it adopts your style or knowledge.
Embedding
Turning text into numbers that capture meaning, so software can find similar or related content.
Vector database
A store of embeddings that lets an AI quickly find the most relevant information to answer a question.
Agent
An AI that can take actions, use tools, search, send, update, not just chat. It works toward a goal, step by step.
Context window
How much text a model can consider at once. Bigger windows mean it can 'remember' more of a conversation or document.
Inference
The act of an AI producing an answer. 'Running inference' just means using the model.
Training data
The huge body of text a model learned from. It shapes what the model knows and how it writes.
Multimodal
A model that handles more than text, images, audio or video too.
Temperature
A setting for randomness. Low is focused and predictable; high is more creative and varied.
Generative AI
AI that creates new content, text, images, code, rather than just classifying or sorting.
Guardrails
Rules and checks that keep an AI's output safe, accurate and on-brand.
API
The plumbing that lets your systems talk to an AI (or any software) automatically.
Chatbot vs agent
A chatbot answers questions; an agent gets things done by taking actions on your behalf.
GPT
A family of language models. 'Transformer' is the underlying architecture most modern AI uses.
Trusted sources
Straight from the people building the models and the standards.
- Anthropic documentation — the makers of Claude, on how their models work.
- OpenAI documentation — reference for GPT models and the API.
- Google AI — research and tools behind Gemini.
- NIST AI — the US standards body on trustworthy AI.
Putting AI to work in your business? See our AI automation and AI agents services.