Give AI Examples – Not Just Instructions
One of the simplest ways to improve the quality of AI’s output is to show it what good looks like.
AI can interpret an instruction in many different ways. Providing an example removes much of that ambiguity.
Instead of:
“Write a short executive summary of this report.”
Try:
“Write a short executive summary of this report for the Board. Keep it under 150 words and focus on business impact, financial implications, major risks and decisions required.
Use this style:
Sales increased 12% during Q2, driven primarily by growth in the SME segment. However, margins declined by 3% following higher acquisition costs. Management recommends reviewing pricing before further expansion.”
You’ve now given AI both an instruction and an example of the standard you expect.
Why this works
Examples help AI understand things that can be difficult to describe precisely:
Tone — formal, conversational, persuasive or concise.
Structure — how information should be organised.
Level of detail — executive overview or detailed analysis.
Format — report, table, email, briefing or recommendation.
Quality — what you consider a good answer.
This is particularly useful for repeatable business tasks.
If you regularly use AI to produce management reports, meeting summaries, customer communications or analysis, give it one or two examples of previously successful outputs.
Take It One Step Further
Don’t just provide an example. Explain what you like about it:
“Use the example below as your guide. Notice that it starts with the business outcome, uses numbers where available, highlights the principal risk and finishes with the decision required. Follow the same approach with the new information.”
Now AI isn’t simply copying the format.
You’re helping it understand why the example is good.
The Executive Habit
Think of it like briefing a new member of your team.
You could spend ten minutes explaining exactly how you want a report prepared.
Or you could say:
“Here’s the report we produced last month. Use this as the standard.”
The same principle works remarkably well with AI.
Executive Takeaway
If you want consistently better AI output, don’t just describe what good looks like – show it.