Because of the poor performance and the expensive license cost, traditional relational database management systems are no longer good choices for processing huge amount of data. Grid computing is replacing the place of RDBMS in data processing. Traditionally a workflow is generated by data experts, which is time consuming, labor intensive and error prone. More over, it becomes the bottleneck of the overall performance of data processing in the grid architecture. This paper proposes a multi-layer workflow automation strategy that can automatically generate a workflow from a business language. A prototype has been implemented and a simulation has been designed.