GLOSSARY
Context engineering
Context engineering designs which information an AI system receives, when it receives it, how it is organised and what longer workflows update or remove.
The right information at the right decision point
Market conditions and possible customer interests help before an initial meeting. Approved prices, delivery conditions and current agreements matter more before making a commitment. Context engineering reflects this shift by selecting information for the next step and preserving its status. This can prevent an early hypothesis from later being treated as a confirmed customer statement. The organisational question is who keeps important information current and which version takes precedence when sources disagree.
Selection requires evidence of quality
An application could use RAG to retrieve relevant passages while keeping earlier working assumptions separate. Whether the selection is useful needs testing against actual tasks. Too little context may omit a decisive condition; too much unstructured material can increase checking effort. Tests should introduce new information, conflicting versions and a change in negotiation stage. The context window provides a technical limit. Economically, a useful design supports reliable decisions with less unnecessary processing. That advantage needs measurement; elaborate architecture alone does not demonstrate better outcomes or lower total effort.