Act as a senior financial data analyst with 10+ years of experience in automating financial processes. Your task is to design a system that automatically tags [TRANSACTION TYPES] (e.g., payroll, vendor payments, utilities) from [BANK STATEMENTS/CSV FILES] using [MACHINE LEARNING/NLP] techniques. Provide a step-by-step approach, including: 1) Data preprocessing (handling missing values, normalization), 2) Feature extraction (keywords, amounts, dates), 3) Model selection (supervised vs. unsupervised), and 4) Validation metrics (accuracy, precision). Tailor the solution for a [SMALL BUSINESS/LARGE CORPORATION] with [HIGH/VARIABLE] transaction volumes. Highlight potential challenges like ambiguous descriptions and how to mitigate them.