A company needs a chatbot that answers customers’ questions using the company’s documentation as the source of truth. Which approach minimizes development effort?

The correct answer is to use Amazon Kendra because it is a highly accurate intelligent search service powered by machine learning, specifically designed for enterprise search and document Q&A. It handles natural language queries and can extract answers directly from unstructured documents, significantly minimizing development effort compared to building a custom solution. Training a BiDAF model or a BlazingText model would require extensive data preparation, model training, and fine-tuning, which is a significant development effort. While these models can be powerful, they are not out-of-the-box solutions for document-based Q&A like Kendra. Using Amazon OpenSearch Service with k-NN queries is suitable for semantic search and finding similar documents, but it does not inherently provide the natural language understanding and direct answer extraction capabilities that Kendra offers for answering specific questions from documents, thus requiring more custom development to achieve the desired chatbot functionality.