Our first post on backlog refinement framed an AI-generated story-point estimate as a useful second opinion – something to compare against your team’s own instinct. A thoughtful pushback we heard directly from an instructor in a recent cohort deserves its own space, because it complicates that advice in a genuinely useful way.
The Counterpoint
Here’s the concern, stated plainly: if the goal of refinement were simply to produce an accurate number, asking AI for an estimate would be a fine shortcut. But that’s often not actually the goal. The real value of a grooming session is the conversation the team has while arriving at an estimate together – the moment someone says “wait, I didn’t realize this touched the payment system too” or “I thought we’d already handled that edge case.” Skip straight to an AI-generated number, and you get a number, but you lose the conversation that number was supposed to produce.
Reconciling the Two Views
Both things are true at once, depending on what you actually need in the moment. If you’re triaging a huge, unrefined backlog and need rough, directional sizing just to get a sense of scale, an AI estimate as a quick second opinion is genuinely useful – nobody’s expecting deep team alignment on 200 untouched backlog items. But if the refinement session’s real purpose is team alignment – making sure everyone shares the same understanding of what’s being built and why it’s harder or easier than it looks – shortcutting straight to a number defeats the actual point of the exercise.
A Practical Rule of Thumb
Use AI to prepare for the conversation, not to replace it. The dependency-finding, redundancy-flagging, and risk-surfacing techniques from our original refinement post are exactly the kind of AI assistance that makes the human conversation richer – the team walks in already aware of a hidden dependency, and spends their time discussing what to do about it rather than discovering it live. Asking AI directly for the story-point number is the one piece worth reserving for situations where the conversation itself isn’t the goal.
Try This This Week
At your next refinement session, use AI only for the preparation work beforehand – flagging dependencies, redundancies, and ambiguous stories – and deliberately skip asking it for a story-point estimate. Let your team size the work the old-fashioned way, together. Notice whether the conversation feels different, richer, or more revealing than sessions where a number – AI-generated or otherwise – was already on the table before the discussion started.
Keep going:
- Ready to go deeper? Check out our upcoming AI courses for hands-on classes with real practice, not just reading.
- Want more like this? Subscribe to our YouTube channel for walkthroughs and demos that pair well with these posts.
- Have questions or want to talk it through? Reach us directly at hello@rocketninesolutions.com.
- Know a colleague wrestling with the same challenge? Send them this post — half the value of this series comes from the people who’ve already tried it sharing what worked.
A Note on How This Was Written: This post was developed collaboratively between Rocket Nine’s training team and Claude, Anthropic’s AI assistant – drawing on real material from our AI for Product Owners class and anonymized insights from the Product Owners who’ve taken it. We think that’s fitting: the same thoughtful human-AI collaboration we teach in class is exactly how we approached writing about it.
© 2026 Rocket Nine Solutions Inc. All rights reserved.