AI and agents
Large language model (LLM)
A large language model (LLM) is an AI model trained on very large amounts of text to predict the next token, which lets it read, write, summarise, classify and reason over language.
How it works in practice
GPT from OpenAI, Claude from Anthropic, Gemini from Google and open-weight models such as Llama and Mistral are all LLMs. You send text in, called the prompt, and get text out. Everything else, including agents, chat assistants and document extraction, is built around that exchange.
Choosing a model for a business task is a trade between quality, speed, cost per token and where the data goes. A small, fast model is often fine for classifying emails; a larger model is worth paying for when the task needs careful reasoning. We pick per task, keep the model swappable, and run the build under the client’s own API account so usage and data stay in their name.
A worked example
A 15-person agency uses a small model to tag incoming briefs by service and budget, which costs fractions of a cent each, and a larger model to draft the first proposal outline, where quality matters more than cost.
Examples in this glossary describe typical builds for companies of 1 to 1,000 people. They are not named client results.















