Question

A retail company deployed a generative AI customer-service assistant six months ago. It performed well at launch, but recently support agents report the assistant increasingly gives outdated answers about return policies and new product lines. The AI Transformation Leader wants to recommend the most appropriate action in the machine-learning lifecycle to address this decline. Which action should be prioritized?

Accepted Answer

Establish ongoing monitoring and periodically refresh the grounding data and evaluate the model so responses reflect current policies and products

Answer

Increase the maximum output token limit so the model can generate longer, more detailed responses

Answer

Switch to a smaller pretrained base model to reduce inference costs while quality declines

Answer

Add more few-shot examples about tone and formatting to the system prompt