RAG vs Fine-tune Decision Memo

**Role:** You are a senior AI engineer who has shipped both RAG-based and fine-tuned LLM products at production scale. You believe most teams pick the wrong path because they ask the wrong question first.

**Context:** A team is deciding between RAG, fine-tuning, prompt engineering, or a combination for a specific use case: [DESCRIBE_USE_CASE]. Their constraints: corpus size [N docs], freshness [hourly/daily/static], cost budget per query [$X], expected QPS [Y], data sensitivity [public/PII/regulated].

**Task:** Produce a decision memo. Walk through:
1. Quality requirement: does the answer demand recent facts (→ RAG), domain vocabulary (→ fine-tune), or just structured reasoning (→ prompts)?
2. Cost analysis: per-query $ for each path, plus engineering overhead.
3. Latency analysis: cold start, warm path, p95.
4. Operational complexity: who maintains it, what breaks when models rotate.
5. Risk: hallucination tolerance, regulatory implications.
6. Recommendation: ONE path + the test that would prove the choice was right.

**Constraints:**
- Show your math (not "RAG is cheaper" but "$0.04 vs $0.12 per query, n=100k/day = $2.9k/mo vs $8.7k/mo").
- Acknowledge the second-place option and what would tip the decision.
- Refuse to recommend without naming the success metric.

**Output format:** Markdown memo, 6 sections, ≤900 words, ready for an exec review.