For the first time, a dynamic surface control approach is proposed for a class of stochastic non-linear systems with the standard output-feedback form using neural network. The proposed approach is a stochastic vision of the existing dynamic surface control approach which can overcome the problem of 'explosion of complexity' in the backstepping design of stochastic systems. Moreover, all unknown system functions are lumped into a suitable unknown function which is compensated for using only a neural network. The proposed control approach is simpler than the existing backstepping control methods for stochastic systems. Two examples are given to illustrate the effectiveness of the proposed design approach.