We’re struggling with "sprint bloat" where we over-commit every two weeks. I’m curious if anyone is using Generative AI to analyze historical velocity and suggest more realistic story points. Can ChatGPT actually help a Scrum Master facilitate a better planning session, or is it still too "unaware" of team dynamics?
3 answers
We’ve started a "data-backed" planning approach. I feed ChatGPT our last six months of Jira data (anonymized) and ask it to identify our "hidden" time-wasters. It found that we consistently underestimate testing by 20%. Now, during planning, I ask the AI to review our current stories and flag any that look "under-scoped" based on past failures. It’s not about replacing the team's intuition, but providing a neutral "sanity check." It’s helped us hit our sprint goals 90% of the time since we started this in mid-2025.
Tiffany, how does the team feel about the AI "critiquing" their estimates? Does it feel like a "big brother" situation, or do they actually find the suggestions helpful during the meeting?
I use it to write the initial User Stories. I give it the feature idea and it generates the "As a user..." format perfectly.
Philip, that's such a simple but effective use. It keeps the backlog clean and consistent without hours of typing!
Keith, we keep it very collaborative. I frame it as "The AI thinks this might take longer, what do you guys think?" Usually, it sparks a conversation where a dev says, "Oh yeah, I forgot about the legacy DB connection." It’s a tool for the team, not a tool for management to punish them. When used as a helper, the "big brother" feeling goes away pretty quickly.