This week we voted story points on a ticket an AI agent had already finished.
It made me wonder: when an agent does part of the work, what are we pointing? The effort it would take a person, or the effort it actually took?
If we point it the old way, the numbers grow fast. A team that used to close 50 points a sprint could close 200, and the velocity chart would look great.
I'm just not sure customers would notice 4x more delivery. Several studies show that more PRs don't translate into delivery at the same pace, so the chart might be telling us something that isn't really happening.
If we point only what the human did, velocity stays flat, but then it stops telling us much about capacity when half the work runs in parallel with agents.
I don't have a good answer yet, and I've been thinking about it all week.
For teams already working with agents: how are you handling story points and velocity? Did you change the metric, or change how Scrum itself works with the agent in the loop?