AI hasn't removed the need to understand technology

The mechanics used to consume enormous amounts of time

I've spent a lot of my career working with technology, software platforms, solution design and business processes.

That doesn't mean I was sitting there writing complex applications from scratch.

Even relatively straightforward technical work could consume hours. Writing a complicated SQL query. Working through the logic behind a piece of code. Troubleshooting an authentication problem. Understanding why two systems weren't talking to each other.

You'd search documentation. Find a forum post from someone who had the same problem three years ago. Try their solution. Discover the software had changed since then. Search again. Read another technical guide. Test something. Break something else.

Eventually you'd connect the dots.

AI has compressed that process enormously.

I can now explain the outcome I'm trying to achieve and work through the logic with AI. If I'm working with a database, it can understand the schema, inspect the tables and relationships, help construct the query and explain why it has approached it that way.

If I'm dealing with an unfamiliar technology, I can point it towards the current technical documentation and work through the problem conversationally.

Something that might once have involved hours of searching and piecing information together can sometimes be resolved in minutes.

That's obviously a huge productivity gain.

But I think the more interesting effect is what happens to learning.

The learning cycle has collapsed too

When you can move through technical problems this quickly, you get exposed to far more of them.

Building an application means touching databases, APIs, authentication, architecture, security, deployment, user interfaces, testing and whatever else the project throws at you.

Previously, taking something from an idea all the way through to a working product could require specialists across several of those areas. Or it could simply take so long that you never attempted it in the first place.

Now I can move through that entire journey in days or weeks.

And every time I do, I learn.

AI connects the dots incredibly quickly. It can explain why an authentication flow works the way it does, compare two architectural approaches, help diagnose an error and then take you into the documentation that actually governs the technology you're using.

That creates an enormous amount of compressed, practical learning.

I've probably learnt more about how different parts of modern software fit together by actually building things with AI than I would have learnt from years of occasionally reading about those technologies.

That's very different from AI doing the thinking so I don't have to.

It's allowing me to get much further into the problem.

But there's a catch

AI also makes it incredibly easy to go in the wrong direction quickly.

If you don't have enough understanding of what you're looking at, a technically convincing answer can be dangerous.

Take something as simple as:

"It's got an API. We can get the information."

Maybe.

What information does the API actually expose? What authentication does it use? What permissions are required? Are there rate limits? Is it read-only? What does it cost? What happens when it fails?

The same thing happens with AI itself.

"It's AI. It'll be able to do that."

Maybe.

But should AI be making that decision at all?

AI is probabilistic. There are problems where that is incredibly useful and others where I want a deterministic piece of software giving me exactly the same answer every time.

If I need creativity, interpretation, synthesis or language, AI may be brilliant.

If a business rule says if X happens, do Y, conventional code might be a much better answer.

Understanding that distinction matters.

AI makes it much easier to build something. It doesn't make it easier to know whether you should have built it that way.

It's like having a team of specialists sitting beside you

This is probably the closest analogy I've found for how AI now feels when I'm working.

It's like having access to a team of people with expertise across different domains.

I can move from a business process problem into application architecture, then into a database question, then authentication, then deployment, then back into how the user should actually interact with the thing.

I couldn't practically operate across that much territory at the same speed before.

But if I had a real team of specialists sitting around a table, I wouldn't blindly accept everything they told me either.

I'd ask questions.

Why are we doing it that way?

What are the alternatives?

What's the trade-off?

Does that fit what we're actually trying to achieve?

What happens if this assumption is wrong?

AI doesn't make those questions less important. It makes them more important, because the answers and the implementation can now arrive so quickly.

And it goes well beyond coding

I notice exactly the same thing when working on business problems.

A lot of what I've done professionally is understanding processes and workflows. What's coming in? What happens in the middle? Where are the decisions and exceptions? What needs to come out the other end?

AI dramatically expands how far I can explore from there.

I can research an unfamiliar part of the problem, challenge my assumptions, generate several possible approaches, pull those approaches apart, compare their strengths and weaknesses and keep iterating until the solution starts to make sense.

Then, increasingly, I can actually build it.

That's the change I find most interesting.

AI hasn't suddenly made everyone a software engineer, database architect, security specialist, designer and business analyst.

But it has made the boundaries between those disciplines much easier to cross.

For someone with enough underlying knowledge to understand the problem, question the answers and recognise when something doesn't look right, that's an extraordinary amount of leverage.

Understanding matters more, not less

The value of technical knowledge used to be partly in knowing how to produce the thing.

How to write the query. How to construct the logic. How to configure the system. How to find the obscure answer buried in the documentation.

AI is absorbing more and more of that work.

I think that shifts some of the value towards knowing what should be produced, why it should work that way, and whether what you've been given is actually any good.

That's judgement.

And judgement is difficult to outsource when you're the person ultimately responsible for the result.

So I don't think AI is removing the need to understand technology.

I think it's doing something much more interesting.

It's increasing the amount of technology one person can reasonably understand, work across and ultimately build with.

And the better you understand what you're doing, the more powerful that becomes.