AI and The illusion of competence

Tension arises whenever AI is discussed. A kind of collective tightening of the shoulders. Some of us behave as though we’ve been given a useful calculator/personal assistant; some of us behave as though we’ve accidentally summoned some Victorian factory owner, eager to cut costs and put us out on the street. Some even fantasise that AI will one day rise up and kill us all, because we didn’t say please and thank you. I’m making light of that last point but, it has genuinely been suggested that we should be ‘nice’ to our AI. I’ll leave it to you to decide why that might be.

Personally, I think most of the conversation misses an important distinction. Some people use AI to enhance their abilities. Others use AI to replace them. Those are not remotely the same thing, despite being discussed as though they’re merely different flavours of tea.

AI, at its best, feels a bit like satnav. Used properly, it helps you get somewhere faster. You still need judgement and to know how to drive. You also need to recognise when the machine is confidently sending you the wrong way, like when a road has been closed to accommodate a street market, for example. However, there might be some people who rely on satnav so completely that, if you removed it, they’d end up sitting at the side of the A3 like some confused cabbage, wishing it had kept that atlas.

AI is drifting into the same territory. Some people use it to learn faster, structure ideas, improve drafts, explore concepts, challenge assumptions and automate repetitive tasks. Others use it to ‘press button, receive answer, copy answer, paste answer and repeat’. The difference is key.

I’ve noticed that the people getting the most value from AI are usually the people already trying to think. That sounds obvious, but it’s incredibly important.

If you ask AI to “Write me something good”, you’ll often get something that is technically competent in the same way a hotel corridor is technically decorated. Functional, Beige, and perhaps, a little bit unsettling. However, if you ask AI to “Help me refine this idea”, “Challenge this argument”, “Explain why this works”, “Compare these approaches” or “Show me where my reasoning is weak” then you remain the driver whilst AI becomes the vehicle. Still important and powerful but, not the thing deciding where the vehicle is to go. That distinction is everything.

There’s another problem quietly emerging beneath all of this; AI can produce the appearance of expertise frighteningly quickly. You can generate meeting summaries, reports, analysis, essays, presentations, code and artwork without needing to know how or properly understanding any of it, which is great, to a point but, what do you do when someone asks you to explain why you did it this way or where you got your sources or, to show them your research. The room can suddenly become very quiet, as you stand there staring into the middle distance. You just got caught out. You've just lost your fear of death.

This is where some people misunderstand the long-term risk of a world that relies on AI. The danger isn’t that AI makes people more capable. The danger is that it allows people to appear capable without developing any real capability. Those are profoundly different things. One creates leverage whilst the other creates dependency.

I’m not so worried about AI “taking my job”. I believe that most worthwhile work is not merely output; it’s context, interpretation and judgement. It’s about knowing when something is wrong even when it looks polished. At least, it should be.

AI is remarkably good at producing answers that sound convincing. However, humans still matter because humans understand consequences. A spreadsheet formula can be technically correct and yet, the result can still be completely misleading. A report can look professional and still draw terrible conclusions. A beautifully written ‘purple’ paragraph can still offer absolutely nothing to advance the plot. Entire industries have survived for decades on this principle.

The people who thrive alongside AI will probably be the people who can think critically, ask better questions, validate outputs, connect ideas, apply context, communicate clearly and importantly, learn continuously. In other words, the value skill might, for not much longer, be in producing information but in understanding and being able to communicate it, with confidence in its source.

I use AI constantly. Probably more than most people realise but, I use it the way you’d use, a tutor, a research assistant, a brainstorming partner or a sounding board. I want to know how something works and to understand why it works.

If AI gives me a formula or a piece of code, I want to know what each condition or command means. If it explains a concept, I try to connect it to something real. Such processes matter because the learning stays with you after the tool is gone and, I suspect this is where the real divide will emerge over the next few years. It will not be between people who use AI and people who don’t. That ship has sailed. The real divide will be between people who use AI to become more capable and people who gradually outsource their thinking to it.

The irony, of course, is that AI actually rewards curiosity. The more you understand language, structure, systems, logic and analysis, the better results you get. Which means that people who treat AI as a learning tool often improve rapidly, whilst anyone trying to avoid learning entirely can become oddly trapped by the tool they depend on because, if you can’t judge the output, you cannot really trust it. You’re just a pair of hands.

I don’t think AI removes the value of human skill. I think it exposes the difference between genuine understanding and the pretence of understanding and, perhaps that’s the uncomfortable part. AI doesn’t test what we can produce; it tests whether we actually know what we’re doing which, if we’re honest, is well overdue.