AINUA AI Academy
F2.4 · Foundations

Beginner · 18 min core lesson · 3 min quick note

Give only the context that helps

Filter for relevance, permission and freshness.

Context designPrivacyInformation literacy
Nia and Kio introducing Give only the context that helps
Cold open

Context is seasoning. Too little gives a bland answer; too much turns the prompt into stew nobody can identify.

By the end, you can

select relevant, authorised and current context while removing noise and sensitive detail.

Quick note · 3 min

Useful context changes the answer and is safe for the chosen tool.

Do now: For one prompt, mark every detail keep, remove, anonymise or verify.

Relevant does not automatically mean permitted.
Why it matters

Better context improves fit, but unnecessary data increases confusion, cost and privacy risk.

01

Use the changes-the-answer test

Ask whether the answer would materially change without each detail. Keep audience, location, constraints, definitions and source material that matter.

Remove biography, old drafts and examples that do not affect the task. Summarise long background when exact wording is not required.

Context earns its place by changing the work.
02

Minimise and anonymise

Use the least personal or confidential data necessary. Replace names and identifiers with roles or codes where possible, and use synthetic examples for learning.

Permission depends on the data owner, tool terms, organisation policy and law—not your confidence that the model will be discreet.

03

Label sources and freshness

Separate instructions from source text. Include dates, versions and origin when they affect interpretation.

Tell the model which source is authoritative and what to do when sources disagree or lack an answer.

Clean context makes missing information visible.