Botrytis bunch rot (BBR), one of the most important diseases of wine grapes, is usually quantified in the vineyard by visual estimation of percentage disease severity on individual grape bunches. This method is prone to assessor error and there is a need for a more objective quantification method that is cost-effective and practical. Near infrared (NIR; 800–2690 nm) and mid-infrared (mid-IR; 2510–25, 770 nm) spectroscopy were investigated as alternatives to visual estimation. Partial least squares (PLS) analysis of the NIR and mid-IR spectra from near-ripe grape bunches from Tasmanian vineyards was used to generate prediction models from both raw data and data pre-processed using the Savitzky–Golay derivative. The entire spectral range for each spectral region was analysed first, after which specific spectral ranges were analysed based on their influence on the initial PLS analysis. The spectral range of 1260–1370 nm with Savitzky–Golay smoothing and first derivative pre-processing produced the PLS model with the highest predictive ability in the NIR spectral region, with a ratio of standard error of prediction to standard deviation (RPD) of 2.2. The spectral range of 8760–9520 nm with Savitzky–Golay smoothing and first derivative pre-processing produced the PLS model with the highest predictive ability in the mid-IR spectral region, with a RPD of 1.7. Both methods demonstrated the potential for spectroscopic quantification of BBR. However, further calibration is required to increase the accuracy of these models, particularly at low BBR severities, if they are to be considered suitable for use in the vineyard.