AI Doesn’t Replace Knowing Your Subject. It Multiplies It.

Domain knowledge is becoming the real difference between people who use AI. Everyone has the same tools now. What they get out of them is not the same at all.

Search your symptoms online and you’ll convince yourself it’s the worst case. A doctor reading the same information knows which details matter and which don’t. Same input, completely different conclusion, because one of you knows the subject.

In code, it shows fast

Our students recently presented the code for their systems development projects. One team had five lines of plain-English comments on every block: five on the try, five on the except. At a glance it looked thorough. In practice it showed that the code had been generated rather than understood.

Generated, not understood

try:
    # Try to open the file
    # This attempts to open the file
    # If the file exists, it opens
    # If not, an error may occur
    # The error is handled below
    f = open(path)
except Exception:
    pass

Understood

try:
    f = open(path)
except FileNotFoundError:
    log.warning("No config found, using defaults")
    f = None

The first version comments on everything and handles nothing: any error at all is silently swallowed. The second has no comments, and it’s the better code, because whoever wrote it knew what could actually go wrong and decided what should happen when it does.

Tools that work for you, and tools that work for others

AI can produce a working tool that runs on your own PC with very little effort. For many people that’s genuinely enough. But without knowing the field, things get missed: the inputs you didn’t think of, the setup on somebody else’s machine, packaging it so anyone but you can install it, and keeping it working once others depend on it. Building it gets frustrating at exactly the points where knowledge would have helped. It works for you. It doesn’t become something other people can use.

The difference is visible

Watch a designer use AI, then watch me use it, and the difference is obvious. One of our students built a presentation with AI that I would never have thought of making; the slides alone made you stop and look. That wasn’t the tool. That was taste, and taste comes from knowing the subject.

Outside tech it’s the same story. We’ve all seen the jokes about civil engineers who learned the trade online during Covid.

So beneath the hype, the people who already know their field are pulling further ahead, faster and with better results. AI is widening that gap, not closing it.

The good news

Every specialist was a beginner once. The path hasn’t changed: learn the fundamentals rather than only the prompts, build things yourself and let them break, and get feedback from someone who has done it before.

The only age-old problem is that we all want a shortcut to the top. There isn’t one. But there is a path, and it’s shorter with someone walking it alongside you.

Give it time and keep going. Before you know it, the pieces fall into place.