Ambient intelligent applications require applications to recognise user activity calmly in the background, typically by instrumentation of environments. In contrast, we propose the concept of Cooperative Artefacts (CAs) to instrument single artefacts that cooperate with each other to acquire knowledge about their situation in the world. CAs do not rely on external infrastructure as they implement their architectural components, i.e. perceptual intelligence, domain knowledge and a rule-based inference engine, on embedded devices. We describe the design and implementation of the CA concept on an embedded systems platform and present a case study that demonstrates the potential of the CA approach for activity recognition. In the case study we track surface-based activity of users by augmenting a table and household goods.