Our team is working on a new AI integration and we are struggling to estimate story points. Some tasks are completely unknown. Should we use high-point spikes, or just accept that our velocity will be erratic for a few months? What is the best way to communicate this uncertainty to our stakeholders who want a fixed roadmap?
3 answers
When facing "unknown unknowns," you should never just guess. Instead, use a "Spike"—a time-boxed research task—to gain enough knowledge to estimate the actual work in the following sprint. For the roadmap, communicate in "Themes" and "Ranges" rather than specific dates. Explain to stakeholders that in R&D-heavy projects, the first 3-4 sprints are for discovery. Use a Fibonacci sequence for points; if something is a 13 or 21, it’s too big and needs to be broken down. High uncertainty should lead to smaller, more manageable pieces of work, not larger, vague estimates.
Have you considered using "Confidence Voting" alongside your story pointing to visually show stakeholders how much the team actually trusts their own estimates?
Just remember that story points are relative. Compare the new AI tasks to something you have done before, even if it’s just a rough comparison.
Exactly, Mark. Relative estimation is the core of Agile. Even if it's new, we can usually tell if it's "bigger" or "smaller" than a standard API integration we've done.
We haven't tried Confidence Voting yet, Steven. That sounds like a brilliant way to quantify gut feelings. If the team points a story at an 8 but the confidence is only 2/5, it tells the Product Owner that we might be headed for a bottleneck. I’ll bring this up in our next Planning session to see if it helps align expectations with the reality of the technical hurdles.