Formulating contradictions is a crucial step in Inventive Problem Solving, allowing engineers to abstract technical challenges and apply systematic solution principles. This study examines the use of Large Language Models (LLMs) and Generative AI for automated contradiction formulation from patents. The system processes a .pdf file of a patent in a pipeline to identify Technical Contradictions and classifies it into TRIZ Inventive Principles, creating a structured knowledge representation. The proposed method utilizes Large Language Models (LLMs) for text summarization, Technical Contradiction formulation, and solution classification into TRIZ Inventive Principles. Both open-source (llama, qwen) and proprietary (gpt-4o, claude-sonnet-4) models are evaluated on a set of 20 patent documents, and results are compared with human expert assessment and existing SummaTRIZ approach. This paper shows the potential of conversational, pre-trained Large Language Models (LLMs) to support systematic innovation by automating Technical Contradiction formulation and TRIZ-based solution classification.