Amorphous computing aims at discovering new approaches for programming and controlling vast number of computational entities. So far, most related studies are mainly on the problem of pattern formation. However, there still remains a great challenge: how to enable amorphous entities to handle computationally hard problems? In order to tackle this challenge, we propose to take the energy efficient routing problem in sensor networks as a special topic for study, due to the similar characteristics of sensor nodes and amorphous entities. In the paper, each data packet sent from a source node is carried by an autonomous mobile data agent, which can make its own routing decision based on its local information. Moreover, we propose a novel data routing idea on energy efficient route choosing. Simulation results demonstrate that our routing schemes based on the proposed idea can achieve a better performance of energy load balancing in the network, and a shorter time delay for data agents to travel from a source to a sink than other schemes discussed in the paper. Furthermore, by endowing the data agents with some intelligence, data aggregation can be performed locally and dynamically during their transmissions in sensor networks.