AI & Technical question

In what situations would you explicitly avoid using RAG and choose prompting or fine-tuning instead?

Practice this question out loud. An AI interviewer asks it, follows up like a real interviewer would, and scores your answer. Type or speak.

What this question tests

Technical judgment on selecting the right AI architecture for a problem, showing understanding of RAG's actual tradeoffs versus alternatives.

State RAG's core strength: grounding answers in external, frequently changing or proprietary knowledge the base model was not trained on.
Identify cases where RAG is unnecessary: when the required knowledge is well covered in training data and does not change often, prompting alone suffices.
Identify cases favoring fine tuning: when the need is a consistent style, format, or specialized skill rather than fact lookup.
Note latency and cost tradeoffs, since RAG adds retrieval latency and infrastructure complexity that prompting avoids for simpler tasks.
Note a case where fine tuning is required: when the retrieval corpus is too sensitive or small to build reliable retrieval over.

Correctly distinguishes RAG's purpose from fine tuning's purpose, rather than treating them as interchangeable.
Names concrete cost and latency tradeoffs, since RAG adds real infrastructure overhead simpler prompting avoids.
Gives a specific scenario for each alternative rather than a generic it depends answer.

Treating RAG as always superior to prompting or fine tuning without naming its overhead and limitations.
Confusing fine tuning's actual purpose with knowledge injection, which RAG handles better.

How would you decide between fine tuning and prompting for a task requiring a strict output format?
What would make you combine RAG and fine tuning together?