The Question Engineers Are Asking — and Why It’s Wrong
When junior engineers pull me aside and ask “what is AI going to do to my career?” — I understand the anxiety. It’s a genuine question born from real uncertainty. But it’s the wrong question.
The right question is: what are you going to do with AI?
The framing matters. One question positions you as a subject — something that happens to you. The other positions you as an agent — someone who shapes what comes next. The engineers who will define the next decade have already made that mental shift.
The Leadership Principle Nobody Teaches in CS Programs
The best leaders I’ve worked with operated by a simple, almost counterintuitive rule: if you’re still doing the same thing you were doing 18 months ago, you’ve failed.
Not because the work wasn’t valuable — it was. But because staying fixed in place means you’ve stopped growing. The model is: replace yourself, then go do something bigger. Hand off what you’ve mastered so you can go tackle what you haven’t. That’s how you scale as an individual. That’s how great organizations scale too.
The irony is that most computer science education — good as it is at teaching fundamentals — rarely teaches this. It teaches you to build systems. It rarely teaches you to replace yourself inside them.
The greatest thing my education gave me wasn’t any particular skill. It was the muscle to adapt quickly to unknowns and the confidence to tackle change decisively. AI is the biggest unknown the industry has handed us in a generation. That muscle is exactly what’s needed right now.
This Isn’t the First Time the Stack Has Collapsed
The compression of engineering roles is not new. We used to treat backend and frontend as distinct disciplines — separate teams, separate career tracks, separate identities. Then abstractions matured. Frameworks improved. And the full-stack engineer emerged: one person owning both sides of the experience.
Nobody mourned the loss of strict specialization. We celebrated the gain in ownership and leverage.
AI is doing this again, but more dramatically and faster. We’re entering the era of the Value Engineer — someone who can own not just the code, but the workflow, the measurement, the business outcome. Someone who can take an idea from inception all the way to a customer paying for it.
The question is not whether the stack gets compressed. It will. The question is whether you’re the one doing the compressing — or the one being compressed.
Where AI Adoption Actually Goes Wrong
Most AI adoption conversations in organizations get stuck in a procurement mindset: find a tool, plug it in, declare success. That’s not workflow replacement. That’s decoration.
Real workflow replacement is a discipline with a clear sequence. It starts with an honest question most teams skip entirely: should this workflow be replaced at all? Many shouldn’t — they should be deleted. The existence of a workflow is not sufficient justification to automate it.
For the workflows that survive that cut, the process looks like this:
Find the right tool — and don’t rule out building your own. Sometimes the workflow is specific enough that a custom-built solution outperforms anything off the shelf.
Build efficacy measurement methodology. The engineer who owns the workflow is the best person to build this — not because they’re uniquely technical, but because they understand the edge cases, the failure modes, the exceptions that never make it into the documentation. Measure first, then use that data to tune the tool into shape.
Build safety guardrails. A workflow without guardrails isn’t a replacement — it’s a liability. This step is what gives you control, and control is what makes the replacement sustainable.
Map token usage to revenue generation. This is the final test. Can you draw a straight line from what the AI is consuming to the value it’s producing? If you can’t make that connection, you’re automating noise. The measurement layer isn’t a nice-to-have — it’s what separates a real workflow replacement from an experiment that quietly dies.
The Fear Underneath All of This
There’s a fear I hear often, rarely stated directly: if I automate what I’m known for, what am I?
It’s a reasonable thing to feel. Identity gets tied to craft. Craft gets tied to specific tasks. When those tasks become automatable, it can feel like the identity itself is under threat.
But here’s the reframe: the engineer who automated the workflow didn’t lose ownership of the domain — they expanded it. They went from being a participant in the process to being the architect of the system that runs it. That’s not a smaller role. That’s a much larger one.
There’s a version of this moment where AI shrinks your role. There’s another version where it multiplies your leverage so dramatically that you’re playing an entirely different game.
You are far more valuable as the entire machine than as a single cog inside it.
The Inflection Point Is Now
Every great wave of tooling in software has separated two kinds of engineers: those who used the new tools to expand their surface area, and those who waited to see what would happen to them.
The backend/frontend split resolved in favor of engineers who chose to own more. The full-stack era rewarded those who refused to stay in their lane. The AI era will be no different — except the stakes are higher and the window to move is shorter.
The best engineers have always been trying to replace themselves. The ones who do it first, and do it deliberately, are the ones who end up with the most to build next.