This paper presents multiagent architecture which detects faults in process automation and allocates intelligent algorithms in field device function blocks to solve these faults. This architecture is a FIPA-standard based agent platform and was developed using JADE and Foundation Fieldbus technology. The main objective is to enable problem detection activities independent of user’s intervention. The use of Artificial Neural Network (ANN) based algorithms, enables the agents to find out about problem patterns and to make decisions about which algorithm can be used in which situations. With this we intend to reduce the supervisor intervention to select and implement an appropriate structure of function blocks algorithms. Furthermore these algorithms, when implemented in device function blocks, provide a solution at fieldbus level, reducing data traffic between gateway and device, and speeding up the process of dealing with the problem. An example is demonstrated with a laboratory test process where fault scenarios have been imitated.