The goal is to transcribe the conversation between the customer and the salesperson to improve the customer experience. The main challenges of this project is to separate what the customer says from what the salesperson says and the distance from the salesperson to the microphone. The microphone will not be worn by the salesperson, but several microphones will be located in different parts of the store, at a distance from the speakers. For this reason, the system must be robust to the other sounds produced in the shops such as music, distant voices, car traffic on the street, movement of goods, etc. To overcome this challenge we trained a neural network system (DNN – LSTM) capable of distinguishing between the voices of the speakers of the conversation and the rest of the distant voices or noises. In the training, the DNN learns how noise is mixed with the voice, as well as learning to distinguish the different sounds of a language in order to transcribe them into the appropriate alphabet. In addition, microphone arrays that can locate speakers are used to increase the success rate.