AI is quickly becoming ingrained in the culture of information technology. Like many of you, I started out using it like an advanced Google, and to be fair it’s great for that. It keeps me from having to go to hundreds of sites, distilling down the information I see, trying a few of the commands to resolve an issue and failing until I get it right. I don’t want to discourage that use case either, while it’s not as powerful as some of the other uses it’s definitely helpful.

AI is at its core best when it’s used as a collaborative mentor or as a tool and it all comes down to context engineering, an extension of prompt engineering. AI overall is smart, has access to pretty much all data but it’s missing a key component in decision making and that is the context you are working in. Think of the context as the guardrails that keep the response on track. AI doesn’t innately understand the limitations that you are set up with, so the more information you give it about what you are thinking about and why you are looking at doing it a certain way will give you vastly different results.

Simple request — no context

AI Promptno-context.md
I would like to fix a clicking noise in my car

That prompt gets you a catalog of every clicking noise a car can make. Useful as a pamphlet. Useless if you already know the year, the symptom, and that you can turn a wrench.

Context request

AI Promptwith-context.md
I am a mechanically inclined tinkerer that can typically fix most minor mechanical issues. I am in IT by trade but understand basic mechanics. I have a 2017 Chevrolet Malibu that has a clicking and popping sound coming from the back of my car while I am driving it. When I get out, I hear a pop in the back, this has been happening for a while and doesn’t seem to cause issues when driving. What are some of the common causes of this starting with the most common issues for this model year and make. I would like to also know how to fix it myself if possible and if it makes sense for me to do it or take it to a shop.

That version gets you the Malibu-specific knuckle bushing, the GM bulletin, and a decision about whether greasing it yourself is enough. Same model, different job, because you told it who you are and what “done” looks like.

This is the power of context. Instead of focusing on the generality of the noise that could be about anything, I have told AI what it needs to know to give me a detailed plan on how to actually resolve this issue. Context is everything to get you to the right answer instead of just an answer that is going to be wrong most of the time. You can use this path for minor or major issues. It saves a ton of time and back and forth chatting with the AI Agent.

Tip: The first prompt in the chat is the most important. It makes you think through what you are really asking for. If you don’t know, it can interview you, but it’s better to understand the outcome than to figure it out mid-thread.