AI terms, translated
The vocabulary you keep running into, defined in plain English with a small-business example. No engineering degree required.
AI agent
An AI agent is software that uses an AI model to decide on and carry out a sequence of actions in your tools — reading an email, looking up a record, drafting a reply — rather than just answering a single question.
Read the full definition →AI hallucination
An AI hallucination is when a language model states something false with complete confidence — an invented fact, citation, or detail. It is a known failure mode that good system design works around, not a rare glitch.
Read the full definition →Context window
The context window is how much text an AI model can consider at once — its working memory for a single task. Everything the model needs must fit in it, which shapes how business AI systems are designed.
Read the full definition →Embeddings
Embeddings are lists of numbers that represent the meaning of a piece of text, so that similar ideas end up numerically close together. They are what lets AI search by meaning instead of by exact keywords.
Read the full definition →Fine-tuning
Fine-tuning is additional training that adjusts an existing AI model on your own examples so it picks up your patterns, formats, or voice. It changes how the model behaves, not what it knows about current facts.
Read the full definition →Generative engine optimization (GEO)
Generative engine optimization (GEO) is the practice of making your business easy for AI systems to find, understand, and cite, so you get named in AI-generated answers the way SEO gets you ranked in search results.
Read the full definition →Human-in-the-loop
Human-in-the-loop is a design where AI does the work but a person reviews, corrects, or approves the result before it takes effect — drafts instead of sends, suggestions instead of actions.
Read the full definition →Large language model (LLM)
A large language model is AI software trained on enormous amounts of text that can read, summarize, and write language. It is the engine behind tools like ChatGPT and Claude.
Read the full definition →Prompt engineering
Prompt engineering is the craft of writing the instructions an AI model receives so it produces accurate, consistent, on-policy results, especially when the same task runs thousands of times.
Read the full definition →Retrieval-augmented generation (RAG)
Retrieval-augmented generation (RAG) is a technique where an AI system looks up relevant passages from your own documents first, then writes its answer from what it found, citing the source.
Read the full definition →Structured data (schema markup)
Structured data, often called schema markup, is machine-readable labeling added to your website that states facts about your business — what you do, what you offer, your FAQs — in a format search engines and AI systems parse without guessing.
Read the full definition →Vector database
A vector database stores embeddings — the numeric representations of your text — and finds the closest matches to a query in milliseconds. It is the search index behind document AI and knowledge assistants.
Read the full definition →Heard a term that isn't here?
Ask us directly — decoding vendor pitches is half our job.
Book a free consultation