Academy / Glossary
The artificial intelligence glossary.
Ten terms, enough to read a supplier quote, tell a real solution from a slide deck, and know what you bought.
Language model
A system that predicts the most likely continuation of a sentence based on an enormous volume of text. It does not search a knowledge base, it generates an answer — which is why it sounds convincing even when it is wrong.
Prompt
The instruction you give the model. The quality of the result depends almost entirely on how clearly the instruction states the role, goal, constraints and answer format.
Context
Everything the model sees at once in a single conversation — your instruction, attached documents and the exchange so far. Context is limited, so long discussions lose the earliest information.
Hallucination
An answer that is linguistically correct and factually wrong. Not a defect but a consequence of how the model works — which is why every business-critical output needs a verifiable source.
RAG
A method where the model retrieves your documents before answering and answers strictly on their basis. The standard way to make a chatbot talk about your prices rather than general internet knowledge.
AI agent
A system that not only generates text but executes steps: reads mail, writes a value into a spreadsheet, sends a reply. Unlike an assistant it works without confirmation at each step, so it needs clear limits of authority.
Fine-tuning
Additional training of a model on your data so it picks up your tone and terminology. More expensive and less often needed than it is offered — in most cases RAG solves the same problem more cheaply.
Token
The unit of text by which model usage is measured and billed. Roughly three quarters of a word; longer documents in the context mean higher cost.
AEO / GEO
Optimising content for the answers of AI tools instead of for a list of blue links. The emphasis is on clear structure, factual accuracy and confirmation of your data on external sources.
Personal data in AI
Entering customer or employee data into a public AI tool is processing of personal data and falls under GDPR. Business versions of these tools offer a data processing agreement and exclusion from training.