Prompt Engineering: Why It Matters More Than You Think

You’ve almost certainly heard the phrase prompt engineering.

It’s become one of those expressions that appears in every conversation about artificial intelligence. It sounds technical and perhaps a little intimidating.

It isn’t.

If you’ve ever typed a question into ChatGPT, Copilot, Claude or Gemini, you’ve already been doing it. You may not have called it prompt engineering, but that’s exactly what it is.

The interesting part isn’t that prompts exist. We all know that. The interesting part is why some people seem to get remarkable results while others decide AI isn’t all it’s cracked up to be.

The difference is rarely the AI.

It’s how it’s used.

A better prompt usually leads to a better answer. That means less rewriting, less frustration and less time trying to explain yourself again. If AI is part of your working day, those savings soon add up.

I’ve never really thought about prompt engineering. I think about directing work.

Perhaps that’s because I’ve spent years running programmes and projects. You define the outcome, brief the team, review what comes back, challenge it, refine it and keep going until everyone agrees it’s ready.

Working with AI feels much the same.

Sometimes it’s writing an article. Sometimes it’s analysing a report, comparing products or researching a subject. Each task becomes a small project. AI is part of the delivery team. I’m the stakeholder and the delivery lead.

That also explains why I don’t expect perfection from the first answer. I wouldn’t expect that from a colleague either. We review it, discuss it, make changes and gradually arrive at something better.

Occasionally I even tell it it’s completely lost the plot.

More often than not, that means I haven’t explained the brief as clearly as I thought I had.

People often ask for the perfect prompt. I’m not convinced there is one.

The best results usually come from a conversation. The first prompt starts the work. The next few shape the answer until it does exactly what you need.

Some people like to follow a structure such as CO-STAR, which stands for Context, Objective, Style, Tone, Audience and Response. It’s a useful checklist, particularly when you’re starting out.

After a while, though, you stop thinking about the checklist. You simply learn how to give a better brief.

One habit has made a noticeable difference for me.

When AI produces an excellent result, I don’t just use it and move on. I save the prompt.

Think of it as keeping a favourite recipe. Once you’ve found something that works, why start from scratch next time?

I keep a folder of prompts that consistently produce the results I want. Some write articles. Others summarise meetings, analyse documents, compare products or generate ideas. They become starting points rather than finished products.

When I need something similar, I copy the prompt, change a few details and begin the conversation again.

Over time, that collection becomes surprisingly valuable. You’re not just saving prompts. You’re capturing ways of thinking that have already proved their worth.

There is another benefit that rarely gets mentioned.

Every time AI has to rewrite an answer, start again or work through several versions because the original brief wasn’t clear, it is doing more work than it needed to.

That costs you time. On many AI platforms it also counts towards your usage allowance or increases costs where pricing is based on consumption. Better prompts usually mean fewer revisions, less processing and a quicker route to the answer you were looking for.

It also makes it less likely that you’ll be greeted with the message none of us enjoys seeing: “You’ve reached your limit. Please try again tomorrow.”

Good prompt engineering isn’t about clever tricks or memorising acronyms.

It’s about communicating clearly, directing work effectively and building a library of prompts that save you time every time you use them.

Most of us already have those skills. We’ve used them with colleagues, consultants and suppliers for years.

Artificial intelligence hasn’t invented a new discipline.

It’s simply given an old one a new name.