Not Losing the Forest for the Trees
How much should you know about AI? The answer, I'd argue, depends entirely on who you are and what you need from it — but for some it'll be more than they think, and for others, the risk is getting so deep you lose the forest for the trees.
If you are someone who just wants AI to help manage your inbox, summarize documents, or plan a trip, you do not need to know programming. But you do need to understand a few basics — for example, understanding what a context window is or why different models handle context differently — otherwise, you are flying blind. On the other end, if you are curious about the tools and have the time, go deep. Run models locally, spend evenings on Hugging Face, and tinker around. That curiosity is valuable. The point is not to warn anyone away from going deep — it is to suggest that wherever you land on that spectrum, you stay oriented toward your actual goal. Do not lose the forest for the trees.
Think about how we approach health literacy. We do not expect everyone to understand cellular biology. But we do expect a baseline: understand that you need enough sleep, you should limit alcohol, move your body, and know the warning signs of something serious. That baseline exists because it is genuinely useful for almost everyone, regardless of how deep they want to go. AI is very similar in this regard.
So what does that baseline actually look like?
Understand what AI is good at. Pattern recognition, summarization, drafting, and data analysis. If your problem fits one of those buckets, AI is probably worth trying. Here are a few simple examples: Have an AI quickly summarize meeting notes so you can share the main points with your team. Generate a shopping list based on your weekly meal plan, or organize your to-do list.
Understand what AI struggles with. It is not a magic solution. It hallucinates. It is forgetful — this is what people mean by “context window”: the more you feed into a conversation, the more it risks losing earlier context. Side note: to mitigate the risk that the model forgets valuable context, just remember to periodically close your chat and start a fresh one.
Understand the basic concepts. By no means do I expect everyone to understand the minutiae of Retrieval Augmented Generation (RAG), but understanding the gist of the technique is incredibly valuable (e.g., in this case, you can think of RAG as a reference library your model can draw from when needed).
Start with your problem, not the tool. Do not go looking for an AI tool to try out. Go to AI when you have a specific problem. That framing keeps you grounded and saves you time.
Beyond that baseline, it is entirely up to you. If your role involves building with AI, integrating it into workflows, or you are simply the kind of person who enjoys pulling things apart to see how they work, go further. There is a lot to explore, and that exploration has real value. But if that is not you, that is completely fine too. You do not need to earn a computer science degree to benefit from these tools.
The goal for everyone is to stay oriented toward what you are actually trying to accomplish. Know your level, stay calibrated, and let that guide how deep you go.
What level are you at, and does it match what you actually need from AI?
Take care,
Emanuel