AI jargon explained: a plain English glossary for business owners
LLM, GPT, hallucination, prompt, token — here is what all the AI jargon actually means.
AI conversations are full of jargon. If you have ever felt lost in a sea of acronyms and technical terms, this is the article for you. Here is every piece of AI vocabulary you are likely to encounter, explained in plain English.
LLM
LLM stands for large language model. It is the technology behind tools like ChatGPT, Claude, and Gemini. An LLM is trained on billions of words of text and learns the patterns of language well enough to generate coherent, useful responses to your questions. When people say "AI" in a business context, they usually mean an LLM.
GPT
GPT stands for Generative Pre-trained Transformer. It is the technical architecture behind OpenAI's models, including GPT-3.5, GPT-4, and their successors. "Generative" means it produces new text rather than just retrieving stored answers. In everyday use, "GPT" is sometimes used loosely to mean any AI chatbot, though strictly it refers to OpenAI's specific models.
Prompt
A prompt is the instruction or question you give to an AI. "Write me an email declining this meeting" is a prompt. "Summarise this document in five bullet points" is a prompt. The quality of the prompt has a huge effect on the quality of the response. Writing better prompts is a learnable skill, and it makes AI significantly more useful.
Token
Tokens are the units AI models use to process text. A token is roughly three to four characters, so a word is usually one or two tokens. This matters because AI models have a limit on how many tokens they can process in one conversation (called the context window). It also affects API pricing, which is charged per token. For most users, tokens are just a background detail.
Hallucination
Hallucination is when an AI confidently states something that is factually wrong. It is not lying. The model has learned to produce plausible-sounding text, and sometimes that produces incorrect information stated with complete confidence. For business use, this is the most important limitation to understand. Always verify anything important that an AI tells you, especially facts, figures, and legal or medical information.
Fine-tuning
Fine-tuning means taking a general-purpose AI model and training it further on specific data to make it better at a particular task or domain. A legal AI might be fine-tuned on thousands of legal documents. A customer service bot might be fine-tuned on your company's knowledge base. Fine-tuning produces a model that performs better for its specific purpose than the general model would.
Open source
Open-source AI models are ones where the underlying model code and weights are made publicly available. Anyone can download and use them without paying. Meta's Llama and Mistral are examples of popular open-source models. The opposite is a closed-source or proprietary model, like GPT-4 or Claude, which are only accessible through the company's official channels.
Closed source
Closed-source models are AI models where the weights and architecture are kept private. You can use the model through an API or interface, but you cannot download or inspect the underlying system. OpenAI's GPT-4 and Anthropic's Claude are closed-source. The company controls access, pricing, and updates.
Multimodal
Multimodal means the AI can work with more than one type of input. A text-only model just reads and writes text. A multimodal model can also process images, audio, or video. GPT-4o, Gemini, and Claude are all multimodal: you can upload a photo and ask questions about it, or describe what is in a screenshot. This is increasingly the standard for leading models.
AI agent
An AI agent is an AI that can take actions, not just answer questions. A basic chatbot responds to what you ask. An agent can browse the web, run code, send emails, book meetings, and chain multiple steps together to complete a longer task. Agents are more powerful but also harder to supervise. The AI agentic space is developing fast and will change how businesses automate work.
Context window
The context window is the amount of text an AI model can hold in its "working memory" at one time. If you paste a very long document, you need a model with a large enough context window to read the whole thing. Older models had small context windows and would "forget" the beginning of a long conversation. Modern models like Claude and Gemini 1.5 have very large context windows, measured in hundreds of thousands of tokens, enough to process entire books in one go.
That covers the main terms you will encounter. The most useful thing to remember: underneath most of the jargon is a single core concept. A very capable text tool that has read an enormous amount of information and can use it to help you work faster.
Explore more on AdaHQ
Everything you need to start using AI in your business.