A retail company currently has operators manually classify customer claims into categories and store each claim record (including its category) in a central database. The company has no data science team and needs a solution that requires no machine learning expertise to automatically assign categories and route incoming claims to queues. Which approach meets these constraints?

The correct answer leverages Amazon Comprehend's custom classification, which is a fully managed natural language processing (NLP) service that allows users to train custom text classification models without requiring machine learning expertise. By exporting the existing labeled claim data (claimlabel and claimtext) into a CSV, it directly provides the necessary training data for Comprehend. The application can then easily integrate with the Comprehend API for predictions. The other options are less suitable because: SageMaker Object2Vec and LDA require more machine learning expertise to configure, train, and evaluate effectively, which contradicts the "no machine learning expertise" constraint. Amazon Textract is for extracting text from documents and images, not for extracting structured data directly from a database. While it could extract text from claim documents if they were unstructured, the problem states the data is already in a database.