Supplying user context and recent actions directly in the prompt is correct because giving the model the user name the last two shows watched and any recent plan changes lets it generate grounded personalized welcome messages without retraining.

By inserting user attributes and recent actions into a prompt the model can reference specific facts and produce timely relevant greetings. This method scales per user and is fast and cost effective because it avoids running a separate fine tuning cycle and instead uses prompt templates or runtime data injection to achieve consistent personalization.

Expanding the response token limit is wrong because increasing the token budget only allows longer outputs and does not add the user specific data needed to shape content.

Switching to a compact model variant is wrong because choosing a smaller or faster model affects cost and latency and does not inherently make outputs more personalized.

Fine-tuning the base model is wrong for this use case because fine tuning changes behavior across many users and requires retraining and maintenance. Fine tuning is useful for broad domain adaptation but it is not the most direct way to produce dynamic per user personalization in real time.

Full AWS Practitioner Certification Question