GlossaryStage 2 of 4 · Context
Context engineering
Context engineering is the practice of deciding what a language model knows before it answers: the instructions, documents, data and history it is given along with the request. In a business it means AI works from the company's own knowledge instead of a blank page, which is the second stage of AI maturity.
What is context engineering in AI and LLMs?
A language model only knows what it was trained on and what it reads in the moment. Context engineering fills that moment with the right material: the relevant decisions, process descriptions, past work and tool results, and leaves out the noise. More text is not better context; the right pieces are.
An example of context engineering
A manager asks AI to draft a reply to a client complaint. Without context, the answer is polite and generic. With context, the model also sees the contract terms, the last three emails and how the team handled a similar case, so the draft fits the client and the company's own rules.
Context engineering vs prompt engineering
Prompt engineering is about how you ask; context engineering is about what the model knows when you ask. Prompts are written per request and per person, while context is built once and shared, usually in an AI knowledge base. That shared base is what makes results stop depending on who is asking.