Judgment Is the Work
AI, responsibility, and modern engineering
Author’s Note:
This is the second part of a two‑part series.
This essay had been sitting unwritten for a long time. A recent LinkedIn post by Scott Hanselman about belonging in software finally nudged me to stop avoiding it.
Lately, I have been hearing a familiar argument again.
Repackaged for a new generation.
That using AI is not real engineering.
That relying on tools means you are skipping the hard part.
That if a system helped you think, you did not really earn the result.
I recognize this voice.
I have heard versions of it my entire career.
I heard it when operations work was dismissed as button‑pushing.
When data work was framed as reporting, not engineering.
When making systems observable was treated as support, not design.
Each time, the boundary of “real engineering” moved just far enough to exclude the work that actually kept things running.
AI is just the latest line people are drawing.
Here is what I want to say to anyone early in their technology journey, or in the middle of a shift, who is hearing that voice and wondering if they belong.
Be careful about who you let define the craft for you.
Engineering has never been about purity.
It has always been about judgment.
It is about knowing what to trust.
What to verify.
What to throw away.
And where the risks actually live.
It is about understanding systems well enough to ask better questions of them. Whether those systems are written by hand, generated by tools, or stitched together from both.
Using AI does not remove responsibility.
It concentrates it.
Someone still has to understand the shape of the problem.
Someone still has to notice when the answer is wrong.
Someone still has to live with the consequences.
That work is not fake.
It is the work.
I see this most clearly when I use AI to explore data. The tool can help me sketch a query, surface a pattern, or suggest a direction. But it cannot tell me when a join subtly changes the meaning of the result. It cannot notice when a clean‑looking answer contradicts how the system actually behaves under load. It cannot feel the quiet wrongness of a graph that technically makes sense, and is still misleading.
That responsibility does not disappear because a tool helped me get there.
It becomes sharper.
If you are drawn to building with these tools, to exploring what they make possible, to understanding their limits and their failure modes, follow that curiosity.
The people telling you it “doesn’t count” are often defending an identity, not a craft.
I am no longer interested in arguing about what counts as real engineering.
I am interested in the work that helps us see clearly.
The work that surfaces patterns.
The work that lets us understand what we have actually built, and what it is doing to the people inside it.
That has been my work for a long time now.
It still is.
Alison + Wiggins