Better prompts are just better briefs
Most professionals overthink prompting.
They assume they need a special formula, a technical structure, or a library of magic phrases. OpenAI’s own guidance is much simpler: tell ChatGPT what you want to know, make, or change. Start in your own words, review the answer, then steer it with follow-up messages.
For everyday work, a short prompt is often enough. But for work that matters, give ChatGPT the same things you would give a colleague: the goal, the context, the expected output, and any boundaries.
Instead of:
Summarise this report.
Try:
Turn this report into a one-page summary for the finance leadership team. Put the decision points first. Keep all approved budget figures unchanged. Flag anything you cannot verify.
That is not “prompt engineering”. It is a better brief.
The part most people miss is boundaries. If the dates must stay fixed, say so. If ChatGPT should only use the files you provided, say so. If it should prepare a draft but not send anything, say so. These small constraints prevent a lot of cleanup later.
For bigger tasks, especially in ChatGPT Work, treat the prompt like a delegation note. Share the source material, name the audience, define the format, and explain how you will review the result. Ask it to check its own work before finishing.
A useful prompt might look like this:
Use the attached quarterly reports to create a leadership brief and a six-slide presentation.
The audience is the executive team. Lead with the decisions they need to make. Separate reported facts from your analysis. Cite each number to its source file. Before finishing, check that the brief and slides agree.
The good news: your first prompt does not need to be perfect. Prompting is a conversation. If the answer is too long, ask for a shorter version. If the recommendation is buried, ask ChatGPT to move it up. If it used the wrong source, correct it and continue.
For non-technical professionals, that is the main shift.
Do not try to sound like a machine. Write like someone delegating real work.
What makes the below so obviously AI generated?
- “Most professionals overthink prompting” is punchy, but it risks sounding like a familiar newsletter hook.
- “That is not prompt engineering. It is a better brief.” is useful, but slightly too neat.
- The structure is tidy in a way that could still feel engineered.
Final newsletter post
Subject: Better prompts start with better briefs
A lot of people make prompting harder than it needs to be.
They think they need a special formula, technical language, or a list of perfect phrases. OpenAI’s own guidance is much more practical: tell ChatGPT what you want to know, make, or change. Start in your own words, review the answer, then use follow-up messages to shape the result.
For simple tasks, a short prompt is often enough.
For important work, give ChatGPT the same things you would give a capable colleague:
- What you want done
- What information it should use
- What the finished output should look like
- What must not change
Instead of:
Summarise this report.
Try:
Turn this report into a one-page summary for the finance leadership team. Put the decision points first. Keep all approved budget figures unchanged. Flag anything you cannot verify.
That second version gives the work a shape. It tells ChatGPT who the reader is, what matters most, and where it needs to be careful.
The most useful part is often the boundary.
If the dates are approved, say they must stay unchanged. If the answer should use only the files you provided, say that. If you want a draft but do not want anything sent or published, make that clear.
Small instructions like these prevent a surprising amount of rework.
For bigger tasks, especially when using ChatGPT Work, write the prompt like a delegation note. Share the source material. Name the audience. Define the format. Explain how you will review the output. Ask ChatGPT to check the result before it finishes.
For example:
Use the attached quarterly reports to create a leadership brief and a six-slide presentation.
The audience is the executive team. Lead with the decisions they need to make. Separate reported facts from your analysis. Cite each number to its source file. Before finishing, check that the brief and slides agree.
The first prompt still does not need to be perfect.
If the answer is too long, ask for a shorter version. If the recommendation is buried, ask ChatGPT to move it up. If it missed a source, add the source and continue. Prompting is not a one-shot test. It is closer to reviewing a first draft with someone who can revise quickly.
For non-technical professionals, that is the useful mindset.
Do not try to talk like an engineer. Give a clear brief, review the result, and keep steering until the work is usable.
Source: OpenAI ChatGPT Learn, Prompting.