Geotechnical engineering relies heavily on expert judgement, contextual understanding, and interpretation of imperfect data. While design codes and numerical modelling tools provide structured processes, many critical decisions in ground engineering depend on tacit reasoning that is rarely written down or formalised. With increasing interest in applying artificial intelligence (AI) to engineering workflows, there is a growing need to articulate the reasoning patterns that experienced practitioners use throughout project delivery. This paper proposes a practitioner-focused framework for codifying geotechnical reasoning, drawing on common design philosophies, uncertainty management strategies, and the limitations of current analysis methods. It highlights why numerical results are often non-unique, why mechanism-based thinking remains essential, and why AI systems must be context-aware to support safe and effective decision-making. The paper concludes that AI should enhance—not replace—engineering judgement, and outlines principles for integrating AI responsibly into geotechnical project delivery.