AI-based transcription may struggle with accurately capturing nuances in spoken language, such as accents, dialects, and emotions. Additionally, it may require human review and editing to ensure accuracy, particularly for complex legal terminology or audio with poor quality.