In Cell-free massive multiple-input multiple-output (CF mMIMO) systems, most works aim to improve the overall performance within the coverage range. It may result in the transmission capacity is not fully utilized in low traffic density areas and is insufficient in high traffic density areas. To overcome this issue, we design a traffic demand-oriented network whose transmission capacity can match with the traffic requirements of each unit area. Firstly, we establish a theoretical model for service capability which is defined as the ratio of achievable sum-throughput to average traffic requirement per unit area. By optimizing the deployment of access points (APs), we achieve on-demand service capability and solve the location of APs with the gradient descent algorithm based on a multiple linear regression model. Finally, numerical results demonstrate that the energy efficiency of the proposed AP deployment method significantly outperforms the benchmark method.