AI Feature Prioritization for ML Teams

Most AI feature backlogs get prioritized using RICE or MoSCoW — frameworks designed for deterministic products, not probabilistic ML features. This template introduces five scoring dimensions built for AI PMs: customer pain severity, data availability, model feasibility, customer impact, and validation timeline. The data dependency gate locks any feature requiring new data instrumentation to On Hold, preventing wasted engineering effort. Built from Prerna Kaul's framework at Amazon Alexa AI.