AI Foundations
Search, retrieval, and generation
Learn the difference between finding existing information and generating a new answer.
4 min readFoundations
Workplace example
A policy answer with a source boundary
If you ask AI to answer a question from one approved policy, the safest instruction is to use only that policy and say when the answer is not present. That protects you from invented interpretations.
What this means
- •Search and retrieval find or bring back existing information. Generative AI creates or transforms an answer from available context and learned patterns.
- •Some tools combine both. They may retrieve a document and then generate a summary, answer, or recommendation from it.
- •The key question is: what source did the answer use, and did the tool stay within that source?
Why it matters
- •People can over-trust generated answers because they look like search results.
- •Generated text can include plausible details that were not in the source material.
- •Clear source boundaries make AI-assisted work easier to review.
Common mistakes
- •Assuming a generated answer came from a verified source.
- •Letting AI fill gaps in a policy answer with general knowledge.
- •Ignoring differences between two tools because both sound authoritative.
What good judgement looks like
- •Ask what information the AI used.
- •Use approved source material where accuracy matters.
- •Treat missing source evidence as a reason to check, not a reason to guess.
Try this at work
- •Ask an AI tool to summarise a public article.
- •Then ask it to quote the lines that support each key point.
- •Compare the summary against the source before sharing it.
How this helps your reassessment
- •You can distinguish retrieval from generation.
- •You know when to ask AI to stay inside supplied source material.
- •You do not treat complete-looking output as automatically source-backed.