Johnny Butler

March 13, 2026

If AI is helping write the PR, the PR should show how AI was used, how much effort it took, and what it cost.

At some point, the business conversation stops being “this is cool” and becomes “what is this costing us, what is it saving us, and is it worth it?”
Ultimately, it gets boiled down to the number that matters most: cost.

I’ve started rolling out a lightweight way to show AI usage inside the PR itself.

Not prompt theatre.
Not vague “AI helped a bit.”
Actual review-friendly signals on the change:

preflight recommendation before the slice starts
pre-flight-recommend.png

baseline snapshot taken before implementation
delta-based AI usage in the PR
pr.png

estimated tokens, cost, model, and reasoning effort

The goal is simple:

If we’re serious about AI becoming part of the software delivery workflow, we should be able to show how it was used, not just the final diff.

For me, this is part of a bigger pattern:

spec → implementation → evidence

The PR should not just contain code.
It should contain enough context to review the slice properly:
what was attempted, what guardrails applied, what was verified, and what the AI usage roughly looked like.

A few things I like about this approach:

portable across repos
lightweight enough to actually use
keeps the PR numeric and concise
encourages preflight thinking before implementation starts
makes cost and effort more visible without turning the workflow into admin

It also reinforces something I keep coming back to:

AI usage needs operational structure.

If you want agents shipping production changes, you need more than “here’s the diff”.
You need repeatable workflow, evidence, and feedback loops.

This version uses:

a preflight step
a baseline snapshot
a small helper script
an AI Usage PR section
repo-specific heuristics docs so recommendations can improve over time

pr.png


It’s still an estimate, not perfect accounting.

But that’s enough to be useful:
for cost visibility, for reviewing scope, and for improving how each repo uses AI over time.

This is the kind of practical workflow layer I think teams will need more of.

Not just “AI wrote code.”
More like:

AI contributed to this slice, and here’s the evidence around it.