I used to think the productivity gain from AI was pretty straightforward:
If coding takes 25 minutes with AI instead of 2 hours without it, that’s a big productivity improvement.
But I’m realizing that’s measuring the wrong boundary.
I used to think the productivity gain from AI was pretty straightforward:
If coding takes 25 minutes with AI instead of 2 hours without it, that’s a big productivity improvement.
But I’m realizing that’s measuring the wrong boundary.
Workday published a breakdown that shows this well. Imagine an engineer’s work normally looks like this:
Implementation: 120 min
Testing: 30 min
Code review: 30 min
Rework: 20 min
Now give them an AI coding agent and implementation drops to 25 minutes. Great!
But inspecting the generated code takes 35 minutes. Code review increases to 45 and rework increases to 40.
The implementation step became almost 80% faster, but the end-to-end workflow only improved by about 11%.
AI didn’t remove all that work. Some of it moved downstream into verification, review, and rework.
So now I don’t ask how much faster the task became with AI. I ask how much faster “useful work” moved through the whole workflow, without more defects, rework, or incidents.
I’m not saying that speed doesn’t matter. I’m saying that local speed is only meaningful if it produces an end-to-end improvement.
Loading comments…
Comments aren’t loading right now.
The post is unaffected; the comment service didn’t respond. Try again in a moment, or come back later.