GLOSSARY
Context
Context comprises the information available to an AI system for its current task, including instructions, conversation history, documents and results returned by tools.
The same statement can have different meanings
“The price is too high” is difficult to interpret without background. It could reflect a budget limit, an internal comparison, uncertainty about value or a negotiating tactic. Information about decision authority, alternatives and previous commitments therefore changes the available responses. An AI assistant needs to distinguish information that is actually available from an interpretation that is merely plausible. Without that distinction, an assumption can quietly become the basis of the next recommendation. Subsequent steps then depend on an uncertain premise.
Relevance matters more than a feeling of completeness
Useful context contains information capable of changing a decision, with its origin and freshness. More documents do not automatically improve understanding. Conflicting versions and irrelevant details may make checking harder. Context engineering organises selection and updating; evidence helps separate assertions from supported facts. A practical preparation question is: which missing fact would make a different option preferable? This limits unnecessary research and shows when asking the customer is more useful than generating another simulated scenario.