Many tools such as mathematical models, intelligent systems etc. are used for controlling the blast furnace process. It has been shown that application of neural networks is also efficient. Large amount of information without clear analytic dependence is noted in case of blast furnace heat. Such information can be processed by means of neural robot. Excel Neural Package is used for training the neural robot. Number of analyzed factors is 24. Output data such as blast furnace productivity, cast iron quality etc. The information is transferred to 23 neurons. Information for training the neural robot has been presented. Efficiency of the neural robot has been shown.