“Large Language Models currently have a low success rate in performing complex, cross-disciplinary, and cross-asset thematic analysis.”