Software architecture is a key factor to scale up Agile Software Development (ASD) in large software-intensive systems. Currently, software architectures are more often approached through mechanisms that enable to incrementally design and evolve software architectures (aka. agile architecting). Agile architecting should be a light-weight decision-making process, which could be achieved by providing knowledge to assist agile architects in reasoning about changes. This paper presents the novel solution of using change-impact knowledge as the main driver for agile architecting. The solution consists of a Change Impact Analysis technique and a set of models to assist agile architects in the change (decision-making) process by retrieving the change-impact architectural knowledge resulting from adding or changing features iteration after iteration. To validate our approach, we have put our solution into practice by running a project of a metering management system in electric power networks in an i-smart software factory.