Question

A global logistics company wants to deploy a customer-facing assistant that answers shipment questions. The relevant information is spread across a real-time tracking database, a policy document library, and a frequently updated pricing spreadsheet. Leadership insists that answers must reflect the latest data at the moment a customer asks and must never invent shipment statuses. Which approach best meets these business requirements?

Accepted Answer

Implement retrieval-augmented generation that queries the live systems at request time and grounds responses in the retrieved records

Answer

Fine-tune a base model monthly on exports from all three systems so the model memorizes the data

Answer

Increase the model's temperature setting so it can generate more comprehensive shipment answers

Answer

Use a large pretrained model with an expanded context window and paste static documentation into every prompt