Azure AI Language Custom Question Answering is the correct option.

Azure AI Language Custom Question Answering is designed to ingest manuals and other documents and to build a knowledge base that answers user questions directly from those files. The service provides document ingestion, retrieval, and answer generation with source attribution so you can deploy a document Q and A chatbot with less custom development work and lower engineering cost.

Azure AI Language Custom Question Answering includes features for handling PDFs and other common file formats, automatic chunking and indexing, and a managed retrieval workflow that reduces the need to build and maintain your own vector store and retrieval pipeline. That combination makes it the fastest path to a production chatbot over product manuals.

Azure OpenAI is not the best choice because it provides general purpose models but it does not include a built in document ingestion and knowledge base pipeline. Using it would require you to build and host the retrieval, indexing, and prompt integration yourself which increases development effort and cost.

Azure Cognitive Search can index documents and power retrieval but it is primarily a search and indexing service. It does not natively provide the managed question answering and generative answer features that simplify building a chatbot that uses manuals unless you combine it with additional services and custom code.

Azure AI Language Conversational Language Understanding is focused on understanding intents and managing conversational flows and it is not intended to ingest large document collections and produce document based answers. That makes it the wrong fit for a manuals driven Q and A bot.

Full AWS Practitioner Certification Question