Cutting forces are small, and in many cases insignificant, compared with noise during the micro-machining of many non-metals. The Neural-Network-based Periodic Tool Inspector (N²PTI) is introduced to evaluate tool condition periodically on a test piece during the machining of non-metal workpieces. The cutting forces are measured when a slot is being cut on the test piece and the neural network estimates the tool life from the variation of the feed- and thrust-direction cutting forces. The performances of three encoding methods (force variation, segmental averaging and wavelet transformations) and two neural networks [backpropagation (BP) and probabilistic neural network (PNN)] are compared. The advantages of N²PTI are simplicity, low cost, reliability and simple computational requirements.