What is an AI Prompt? The 4 parts of a good one

Hey, my name is Mahan. I'm the CMO of Nairon. My name is Luka. I'm the CEO of Nairon. Luka, in this video, let's break down prompts. What they are, how we can engineer them to give us the best possible outcome. When working with a prompt, a prompt is essentially the input that we give an LLM to then give us the output that we're looking for. So, it's really important, and this is essentially the piece where we have a lot of influence on. And so the way that I like to look at it is a really good prompt is a really good brief that you give for example to a freelancer.

So a example of a bad brief that you would give a freelancer is design a logo for my company. The problem is that you'll probably have to go back and forth with the freelancer and asking for the most basic questions like what's your uh style? What's your brand voice?

And there's probably going to be a lot of frustrations on both sides. And so an LLM is nothing different. It is essentially that freelancer. Um it's also a freelancer that never gets annoyed with you. And so if you can give it that very good structured brief ahead of time, um you'll actually get a lot further in your sessions. And so a good brief to a freelancer and to an LLM would be, hey, we are a B2B SAS company uh specifically for finance teams. This is exactly what our branding looks like.

Navy, blue, modern. These are our logos that we like to use. And specifically, this is what I need, X, Y, and Z. So, what are the core ingredients that actually make for a really good prompt? There are four very important ingredients when building out a good prompt. Number one, and this is probably what a lot of people are doing, is context. Um, saying something like, I run a B2B SAS company selling X, Y, and Z product. And that's essentially it what we give it in terms of context.

However, a very good prompt gets into specifically who, what, and when this is about. It also gives an example. If you can give it some copy that you used perhaps in the past that you like, that's also really really good to do.

Also, putting constraints. So, actually saying keep this under 100 words, uh, keep this friendly. Those are really good when um when moon building out a prompt. You could say that iterating over the prompt is the fifth also, but we'll cover that in a bit.

Okay, that's great. And I think most people when prompting an LLM, just from uh memory or muscle memory, you could say, they actually prompted like they would a search engine like Google, I think that's what makes for more so like mediocre results. Correct.

Yeah, definitely. So one very big use case of LLMs is obviously doing similar search type queries like that you would do on Google. So in this case if we were to ask Google what are the best running shoes in 2026 essentially would just do keyword matches across a very big databases database of pages and it would give you a final answer very conclusive.

However the the LLM is much better when you want a more customized solution. So saying something like, "Hey, look, I'm flatfooted. I'm 200 lb. Um, and I'm looking for a daily trainer." Uh, that is much better for an LLM to really go through all the options and identify which one is best for your certain use case. And so context is extremely important in this case.

And to that end, I think it's much more difficult to get an ideal outcome if you don't start the conversation off on the right foot with an LLM with the correct prompt. Yeah, 100%. I that's one of the first things that we see when we consult with clients is um that they are trying to correct course in the middle of a session instead of just clearing the session starting again with a much better prompt. And so what we recommend is if you're not going in the right direction, clear your sessions, perhaps take even pieces of that chat that you had um and construct the best prompt possible to really get you in the right direction. Now, I know we're going to be talking about context windows in the next video, of course.

Um, but I think this is a good introduction to prompting. Yes, we'll do a deep dive into context windows and how those affect the outcome of uh of what the LLM gives you as well. In the meantime, you can subscribe to our newsletter. Links down in the description below. You can also follow us on LinkedIn. We're very, very active on there. Until next time, we'll see you in the next video.