We All Became Architects

Last week, Zone01 Kisumu hosted an AI & Software Engineering mini-conference at Rotary, featuring a panel discussion followed by a keynote from Amariah Abishai, Founder of Atlarix.

I helped organize the session, but I left with a question that I think many developers are quietly asking:

If AI can write the code, what am I actually here to do?

The panel: what happens to software engineering?

The panel explored questions many developers are facing:

• Is AI making us better engineers or just better at prompting?
• How do we trust code we didn't write line by line?
• If AI can write code for junior developers, what do they actually need to learn?

There wasn't one perfect answer.

But one thing became clear: the fundamentals still matter. The workflow is what has changed.

We all became architects

Amariah's keynote built on this idea with a line that stayed with me:

"Nobody sent a memo, but the job quietly changed."

The shift can be thought of as moving from bricklayer to architect.

Before, a lot of our value came from writing and implementing the code ourselves.

Now, AI can help lay many of those bricks.

Our responsibility increasingly becomes deciding:

1. What to build : Is this actually the right problem?
2. What not to build : Can we keep the scope focused?
3. How to structure it : How should the system and its components work together?
4. When it's wrong : Can we recognize when AI-generated code is plausible but incorrect?

The machine can help lay the bricks.

We still own the blueprint.

Does AI mean we need to learn less?

This is especially important for junior developers.

There is a real temptation to let AI generate everything and simply move on. But if you don't understand what the code is doing, eventually you'll reach a problem you can't prompt your way out of.

Using AI shouldn't mean:

AI writes → I paste → I move on.

It should look more like:

I think → AI assists → I understand → I verify → I improve.

The faster AI becomes at generating code, the more important it becomes to understand the code well enough to question it.

Verify, don't trust

One of my biggest takeaways from the session was simple:

AI-generated code still needs engineering judgment.

A model can produce code that looks correct, compiles successfully and still contains a serious flaw.

That's why fundamentals, testing, code reviews and understanding system architecture remain important.

AI can accelerate implementation.

It doesn't remove responsibility.

Where does that leave us?

The tools are becoming increasingly accessible. Anyone can generate a prototype.

So perhaps the scarce skill isn't simply "Can you write code?"

It's:

What is worth building, how should it be built, and how do you know it's good?

Maybe that's what it means to become an architect.

Nobody officially gave us the title.

The tools changed, and the role changed with them.