“Professional but warm” and “concise but thorough” mean something slightly different to every model and context. You can keep refining the description, or you can show the target.
Examples work like photographs. They demonstrate vocabulary, structure, register, level of detail, evidence, and what you deliberately leave out, all at once.
What counts as an example
Your own writing, a previous AI output that worked, a reference piece, a template, a report, a screenshot of a layout, or a table can all narrow the interpretive space. The example does not need to be perfect. It needs to be close enough for the AI to recognise the pattern.
Positive and negative examples together are even stronger: “Write in this style, not like this.” The positive example sets the target and the negative example removes a likely wrong direction.
Build a small example library and add the best examples to the relevant project instructions. When you provide two or three worked examples before a task, you are using few-shot prompting, one of the most effective professional techniques that requires no technical knowledge.
Before an important recurring task, ask: do I have an example of what done looks like? If yes, include it.
This is part of the Parenting Your AI series, a practitioner's guide to building AI skills that are safe, effective, and worth trusting. Written from inside enterprise AI systems by someone who has spent years diagnosing what goes wrong when AI meets real work at scale.
Read the full series at KnowledgeManagement.ie