Part 1: The Art (and Science) of Talking to Machines

So, What Is Prompt Engineering?

If you’ve ever typed a question into ChatGPT or any other AI model and been amazed by the response — you’ve already done some prompt engineering. Maybe without realizing it.

But here’s the catch: these models don’t actually understand what you want. They respond based on how you ask. Sometimes, changing just a few words or adding an example can turn a confusing answer into something spot-on.

That’s what prompt engineering is all about — writing your instructions in a way that helps the model give better, clearer, and more useful answers.

For example, asking:

Summarize this paragraph.

...might give you something vague.

Summarize the key action points from this email in 3 bullet points.

...gives you something much more helpful.

In this series, we’ll show you how small changes like these can make a big difference — whether you’re building an AI tool, working with a chatbot, or just experimenting. The goal is to make your prompts clear, consistent, and reliable.

And don’t worry — you won’t need a deep background in AI to follow along, at least not at the start. Just some curiosity and a few real examples. But hey, even after that, if you have the interest, I am sure you'll pick it up. I am learning along the way, so I don't see why it'd be difficult for you. Let’s get started!

Why Prompt Engineering Matters

On the surface, using an AI model feels easy — you type a question, and it answers. But if you’ve ever gotten a weird or vague response, you know it’s not always that simple.

The trick is this: AI models don’t really understand your intent. They respond based on the exact words and structure you give them. That’s why small changes in how you ask can make a big difference in the answer you get.

Let’s look at a quick example. You might try this:

Summarize this email.

The model might give you a long, general summary. But now if you try:

Summarize the key action items from this email in bullet points.

That’s more specific — and the model responds with a cleaner, more useful output.

This is where prompt engineering comes in. It gives you more control. Instead of hoping the model understands you, you guide it — clearly and intentionally. This becomes especially important when you’re building real tools: writing emails, answering support questions, generating code, or pulling insights from documents.

The better your prompt, the better your results — and the fewer surprises you’ll get. As you build more complex systems, prompt design becomes part of your product’s logic. It affects how reliable, safe, and helpful your AI becomes.

That’s why we’re diving into it in this series.

What Influences LLM Output?

AI models like ChatGPT don’t actually “think.” They don’t understand you the way a person would. What they do is predict — one word at a time — based on patterns they’ve seen during training.

That means your prompt isn’t just a question. It’s the starting point that guides what the model predicts next. And the way you write that prompt has a huge impact on the result you get.

Here are some of the key things that affect how an LLM responds:

1. How you word the prompt. The model doesn’t actually “know” your goal — it guesses based on what you write. For example:

What are the pros and cons of remote work?

…will give you a more general answer, while this:

List 3 advantages and 3 disadvantages of remote work in bullet points.

…is clearer, and helps the model format the output the way you want.

2. Whether you include examples. Sometimes giving the model one or two examples helps a lot. This is called “few-shot prompting.” Without examples (“zero-shot”), your instructions have to be extra clear.

3. The system instructions behind the scenes. Some apps send hidden instructions to set the tone or behavior of the model — like “respond like a helpful tutor” or “be concise and formal.” These affect the model across the whole session. For example, Webex AI Agent has a system end guardrails.

4. Model settings like temperature and max tokens. Think of temperature as the creativity dial — higher means more random, lower means more focused. Max tokens controls how long the model’s output can be.

5. How much context the model can remember. LLMs have a memory window (called a context window). If your prompt is too long, older parts might be ignored. It helps to put the most important instructions near the end or keep things short.

When you write a good prompt, you’re not just asking — you’re guiding the model’s next move. That’s what makes prompt engineering such a useful skill.

From Tricks to Technique: How Prompting Grew Up

When people first started using language models, prompt engineering wasn’t really a thing. You’d try something, see if it worked, and if it did — great. If not, you’d reword it and try again. Most of it was trial and error, with tips being passed around on Twitter and Discord.

A lot of it felt like magic: “If you say it this way, it works better.” But it wasn’t clear why.

As developers began using LLMs in real products — like chatbots, support tools, content generators, and coding assistants — things had to get more reliable. Guesswork wasn’t enough. People needed prompts that could be reused, shared, tested, and improved.

That’s when prompting started to become more of a proper skill. Patterns began to emerge — like giving the model a role (“You are a helpful assistant”), walking it through steps (“Let’s think this through step by step”), or showing examples of what you want.

These techniques weren’t just about getting “nicer” answers. They helped make the model more consistent, especially in tricky cases where you really needed it to follow instructions carefully.

Today, good prompting blends ideas from many fields — writing clear instructions, designing good user experiences, even teaching methods. The best prompt engineers think like builders and teachers at the same time.

So if it feels a bit like art and science combined — that’s because it is. And that’s what makes it so powerful.