Companion videos

Choosing between prompting, RAG, and fine-tuning (and when to combine) — companion videos

This is the decision that teams keep getting wrong — fine-tuning a model when the real problem is retrieval, or stacking RAG when a better prompt would solve it. The article lays out a decision framework; these two IBM Technology explainers give you the same framework with a whiteboard and the comparisons drawn out side by side, which is useful when you are trying to make the same case to a sceptical colleague.

Primary pick

13:10

RAG vs Fine-Tuning vs Prompt Engineering: Optimizing AI Models

IBM Technology

A clear whiteboard pass through all three techniques with their respective costs — retrieval latency, training compute and catastrophic forgetting, the limits of prompt-only solutions — and the combinations that actually make sense in production. The closing example of a legal AI system using all three is almost exactly the article's "when to combine" argument.

What you should get from this: Choose between prompt engineering, RAG, fine-tuning or a combination based on the actual failure mode.

Watch or know first: Know what each technique means at a high level.

AI Expert note: This is a stable decision framework. Use it before buying infrastructure or starting a fine-tune project; most wrong choices come from misdiagnosing freshness, attribution, style or behavior problems.

Open video page

Also worth watching

8:57

RAG vs. Fine Tuning

IBM Technology

Tighter focus on the two techniques teams most often confuse. Goes deeper on data freshness, source attribution, and the inference-time speed argument for fine-tuning. Worth watching if you are specifically trying to argue against an unnecessary fine-tune project.

What you should get from this: Explain when retrieval is the right fix and when fine-tuning may actually help.

Watch or know first: Basic RAG and fine-tuning vocabulary.

AI Expert note: Good for stakeholder conversations because it separates knowledge freshness from model behavior. Still validate current fine-tuning costs, supported models and data-governance constraints before committing.

Open video page