Determining adoption of AI through Git commits

Let's say that you're trying to determine what percentage of your work is using AI - for instance as part of a performance review - how would you go about doing that?

I've written before about how I personally attribute code and content from a Large Language Model. I still very much appreciate myself following the practice, and find it useful - even if others don't want to follow a similar process - as it helps "keep myself honest" with noting where these tools have been in use, and reflecting on where I'm using input from external tools.

Fortunately, because I follow this process, I'm able to answer this question in a fairly straightforward manner.

As a good adopter, I decided that the best way to ask how to do this would be to get Claude Code to determine this for me!

With the help of Claude Sonnet 4.6, the following script was the result:

% git log --since="2025-09-01" --author="Jamie Tanna" --format="%H %cd" --date=format:"%Y-%m" | while read hash month; do
  body=$(git show -s --format="%B" "$hash")
  if echo "$body" | grep -qi "co-authored-by.*\(claude\|gpt\|copilot\)"; then
    echo "$month AI"
  else
    echo "$month human"
  fi
done | sort | uniq -c | awk '{print $2, $3, $1}' | sort

This then ends up (after some ingestion from Claude) with the following table:

Pretty cool, eh?