Yes, the problem of accurately measuring software "productivity" has stymied the entire industry for decades, but people keep trying. It's conceivable that you might be able to get some sort of more-usable metric out of some systematized AI analysis of code changes, which would be pretty ironic.