Ask HN: How should Product managers and data scientists work together?

I'm currently working with multiple engineering teams as a product manager, and while scrum works great with most teams, the type of work we're doing with the data science team doesn't seem to fit well with two-week long sprints. From my end I'm feeling a lack of clarity, user stories don't really make sense in the DS/AI sense, and it doesn't really fit in the whole agile/scrum mindset of delivering incremental value in small batches. I know this is inherently tied to the nature of AI/data science work, which is more probabilistic and hard to predict, vs. "traditional" software development that is more deterministic, and so very much predictable. Was wondering if anyone here has had any experience with alternative methodologies, frameworks or just other ways of managing AI initiatives.