Why Human Engineers Still Matter in the AI Era

The obvious question: isn't AI enough?

Why not just ask AI? No consulting fees, no waiting, instant answers. It's a reasonable thing to wonder, and more people are asking it every day.

These days a single instruction gets you a website, or takes care of file operations on your machine. Video, 3D modeling, images, music — nearly everything we used to file under "information processing" is now covered.

I use AI daily myself. For writing code, for looking up technical details, it has become the first place I turn. In both breadth of knowledge and speed, there are more and more situations where it frankly outperforms a human.

But applying its answers as-is is still risky

That said, copying AI's answers straight into a real project still feels dangerous to me.

AI answers the question in front of it accurately. What it doesn't know is how that answer plays out inside the larger system you're actually working on, or whether it meshes with the constraints specific to your situation. An answer that's correct as a general solution often turns out to rest on assumptions that don't hold in a particular place.

Note: Even when AI's answer is technically correct, applying it without context lets problems accumulate in ways that are hard to notice until later.

Can the answer come from somewhere unrelated?

Maybe I feel this way because I learned things the long way around.

I started studying IT seriously as part of a school curriculum. I eventually specialized in software and graduated with a research project in it, but before getting there I worked through pretty much everything computer-related — software, hardware, and the rest.

To understand why a computer physically works, to grasp it as a set of physical phenomena, I studied electricity, magnetism, electromagnetic waves, light, energy, heat, matter, atoms, electrons, crystals, plasma, and on and on.

And as the groundwork for designing software that controls those phenomena, there was mathematical logic, set theory and relational algebra, graph theory, automata and formal languages, calculus, probability and statistics, optimization theory, linear algebra, trigonometry, mathematical physics, number theory, and more.

More directly, I learned the history of the computer itself from its very beginnings (assembly language was quietly good fun), did exercises in a range of programming languages, and worked through the standard algorithms.

Philosophy and ethics, economics, literature, foreign languages — subjects that look far removed from computers — were required too.

Because of all that, though, I feel like I was shown the whole landscape of knowledge surrounding computing. Honestly, at the time I had my hands full just trying to understand each piece, and it was crammed in so densely that there was no room to wonder what any of it was for.

In fact, it was a good while after graduating before I could see how those pieces connect to one another. But as a result, I think I came to recognize a lot of patterns. Phenomena and structures that hold in worlds seemingly unrelated to computing — I started being able to ask whether they might apply to the problem in front of me.

Of course, none of this is determined by whether you studied it in school. Trial and error on real projects, or crossing over from an entirely different field, builds the same thing just as well.

That kind of range — the perspective that steps back and looks at a problem from just outside the knowledge computing strictly requires. The intuition to pick up a hint from somewhere that appears unrelated. This, I think, remains a human's role.

Two roles that will always require a human

1. Someone who understands the whole picture and puts AI to work on it

Solving a specific problem starts with understanding where it sits in the bigger picture and what is genuinely needed. Only with that understanding can you see what to ask AI, how far to let it run, and where the line is that a human has to judge.

We'll always need people who treat AI as a tool to direct — not an oracle to defer to.

2. Someone who evaluates whether AI's solution actually fits

A solution AI proposes can be technically sound and still leave open the separate question of whether it's genuinely the best fit here. Operating cost, maintainability, the team's skill set, future scalability — someone needs to weigh the factors AI can't fully account for.

Skip that evaluation, and you accumulate systems that run but that nobody can touch, or choices that are faster now and cost more over the long run.

The position I take on

My work centers on simulation with NVIDIA Isaac Sim and ROS2, along with the development of web systems and business tools — and on every project, these two roles are what I keep in mind. Not passing AI's proposals straight through, but understanding the whole picture, putting AI to work within it, and evaluating whether the result truly fits the situation on the ground.

Precisely because AI is advancing this fast, I feel the demand for that position is growing rather than shrinking.

Rule-Based Automation vs. Physical AI — Which One Should You Choose? — Choosing between technologies to fit the situation
Bridging Reality and Simulation: Integrating TM Robot with Isaac Sim via ROS2 — An implementation built around real-world constraints
Strategic Consulting — Get in touch about mapping out the bigger picture
AI Integration — Get in touch about designing and evaluating AI adoption