“A blog post by Anthropic concluded that using AI to generate semantic models can negatively affect overall results by anchoring the AI to a specific use case.”