A language model knows a lot in general and nothing about your situation. The context you provide is often the biggest lever on quality.
What Context Helps
- Purpose: why you need the output and what decision it supports.
- Audience: their knowledge level, role and what they care about.
- Background: relevant facts about your organisation, product or project.
- Constraints: budgets, policies, deadlines, standards.
- Prior work: what you've tried and what didn't work.
- Examples of good output.
Compare
Write a job ad for a data engineer.
versus
Write a job ad for a mid-level data engineer at a 40-person Australian logistics company. The role maintains our Snowflake warehouse and dbt models and builds new pipelines from our transport management system. We value clear communication and offer hybrid work in Brisbane. Tone: friendly, plain English, no buzzwords. About 300 words.
Don't Assume Shared Knowledge
Acronyms, internal project names and "the usual format" mean nothing to the model unless explained.
Relevant, Not Exhaustive
Too much irrelevant context can distract. Include what bears on the task.
Reuse Context
For recurring work, keep a short context block — about your organisation, style and audience — to paste in or place in a system prompt.