Prompting vs RAG vs Fine-Tuning

Learn when to improve prompts, when to add retrieval, and when fine-tuning may be worth considering.

Explanation

Prompting is often the fastest way to shape output style or format.

RAG is best when answers need current, domain-specific, or permissioned knowledge.

Fine-tuning is useful when behavior patterns must be learned consistently and examples are abundant.

Why this topic matters in practice

In generative AI products, the model is only one part of the system. The surrounding workflow determines whether the output is useful, safe, and maintainable. This lesson matters because it helps you connect the idea to tasks such as tutoring, search, copilots, business assistants, and production automation.

Examples

A simple decision helper

The code below is intentionally concise so the underlying pattern stays clear. It focuses on the application logic you can reuse, even if you later switch model providers or deployment environments.

def choose_approach(problem):
    if problem == "needs current knowledge":
        return "Use RAG."
    if problem == "needs consistent style":
        return "Improve prompting first; consider fine-tuning if needed."
    if problem == "formatting issue":
        return "Start with better prompting and validation."
    return "Evaluate the workflow and test alternatives."

for p in ["needs current knowledge", "needs consistent style", "formatting issue"]:
    print(p, "->", choose_approach(p))

How the coding section works

• This logic is simplified, but it reflects a practical decision sequence.
• Many teams overuse fine-tuning when retrieval or prompt design would solve the real issue.
• Choosing the right layer saves cost and complexity.

Implementation advice

When turning this lesson into a real feature, think beyond the code snippet itself. Decide what inputs should be allowed, how you will validate outputs, how you will recover from errors, and how you will measure whether the feature is actually helping users. Those surrounding choices often determine whether an AI feature feels polished or unreliable.