Are you still cataloging man hours?
Are you doing everything in story points?
How do you account for the time it takes the agent(s) to do the work?
This is a dilemma I have at work currently. I can estimate how long it would take me to do something but Claude?
If I were in this situation, what I'd do is to continue to give estimates as I always have before, ignoring the use of genAI, and keep track of how much the accuracy of those predictions have changed. That would give me some real data on which I can begin to adjust my time predictions.
To do it any other way is to just be guessing blindly.