1 min lesson
RAG to ground generation in your context
Show why this lesson detail matters: "Retrieval-augmented generation feeds the model your own context at prompt time."
Step 1 of 3
RAG to ground generation in your contextstop the model from making things up
Retrieval-augmented generation feeds the model your own context at prompt time - product docs, past closed-won notes, the account’s history - so the output is grounded in fact instead of the model’s priors. In GTM, RAG is what makes personalization true: the first line references a real case study because the case study was retrieved and pasted into the prompt, not recalled from training data.
Learn more
Full explanation
When a deterministic rule beats an LLM
When a deterministic rule beats an LLMthe cost/latency/quality triangle
Reaching for an LLM where an if statement would do is its own kind of inexperience. LLMs cost money per call, add latency and are non-deterministic. A rule is free, instant and auditable.
- Job
- Is email domain == company domain?
- Right tool
- Rule
- Why
- Exact string compare; an LLM adds cost and a chance of error
- Job
- Map country to sales region
- Right tool
- Lookup table
- Why
- Fixed mapping; deterministic and free
- Job
- Read a paragraph and judge ICP fit
- Right tool
- LLM
- Why
- Requires reasoning over unstructured text
- Job
- Write a personalized opener from notes
- Right tool
- LLM + RAG
- Why
- Generation grounded in retrieved facts
| Job | Right tool | Why |
|---|---|---|
| Is email domain == company domain? | Rule | Exact string compare; an LLM adds cost and a chance of error |
| Map country to sales region | Lookup table | Fixed mapping; deterministic and free |
| Read a paragraph and judge ICP fit | LLM | Requires reasoning over unstructured text |
| Write a personalized opener from notes | LLM + RAG | Generation grounded in retrieved facts |
Use the LLM where reasoning over messy text is unavoidable; use a rule everywhere else.
An LLM is the wrong tool when a wrong answer is costly and silent. If a misclassification routes a whale to the nurture bin and no one notices, the non-determinism is a liability, not a feature. Gate high-stakes actions behind a rule or a human check and reserve the LLM for low-stakes, high-volume judgment where an occasional miss is cheap.
QYou build an AI column to classify leads by ICP fit. What single design choice most improves its reliability in a pipeline?
Learn more
Optional practice