AI glossary for businesses
Large language model (LLM)
In short
A large language model (LLM) is an AI model trained on very large amounts of text to understand and produce human language. It works by predicting the most likely next piece of text, which lets it answer questions, summarise, translate and draft. LLMs power tools such as ChatGPT, Claude, Google Gemini and Microsoft 365 Copilot.
How an LLM works
You already know a tiny version of this idea: the word suggestions on your phone keyboard. An LLM does the same thing at a vastly larger scale. During training it reads a huge share of public text and learns which pieces of text, called tokens, tend to follow others. After that, developers tune it with human feedback so it follows instructions and answers politely.
When you ask a question, the model does not look the answer up in a database. It writes a reply one token at a time, based on the patterns it learned and on what you gave it in the conversation. That is why answers can be fluent and useful, and also why they can be wrong.
"Large" refers to the size of the model: billions of internal settings, called parameters. Well-known LLM families include OpenAI's GPT models, Anthropic's Claude, Google's Gemini, Meta's Llama and Mistral's models.
In practice for a small business
Consider an online clothing shop that gets about forty customer emails a day about sizes, delivery times and returns. An LLM inside a tool like ChatGPT or Claude can draft a reply for each email, in Greek or English, based on the shop's return policy pasted into the instructions. A staff member reads, edits and sends. The AI for customer service replies page walks through the setup.
A practical tip: test models in Greek, not only in English. Quality in Greek has improved a lot, but it still differs between models and tasks. Give each candidate the same five real emails and compare the drafts side by side.
Why it matters and what to watch out for
Knowledge cut-off: a model knows only what was in its training data, up to a certain date, unless the tool also searches the web or your documents.
Made-up answers: LLMs can produce confident errors, often called hallucinations. Keep a person in the loop for anything customers see.
Cost: through an API you usually pay per token, so long documents and long chats cost more.
Data: check whether the provider uses your inputs for training and where data is stored. If you process personal data at scale, consider a DPIA under GDPR.