Amazon's Last Mile delivery network moves billions of packages a year. When a delivery station misses its performance target, the question isn't “what” happened- it's “why”, and “what” we should do about it. That's the problem you'll solve. You'll own the knowledge and attribution platform that sits underneath all our AI and analytics products. Think of it as the structured brain that makes our AI specific instead of generic. When our AI assistant tells a station manager exactly why productivity dropped 10% and gives them four prioritized actions to close the gap by tomorrow instead of a vague summary - it's because the knowledge layer you built made that precision possible. When a VP reads an auto-generated business review and every claim traces back to a verified source with zero fabrications, it's because the ontology you govern enforced that consistency. This role is different from most PM-T roles because you won't own a surface - a dashboard, an app, a feature. You'll own the depth. The structured knowledge, causal logic, and self-service platform that an entire portfolio of AI products depends on. Your decisions about how metrics are defined, how causal relationships are modeled, and how teams contribute new knowledge will determine whether our AI products are trusted or ignored.

Build and govern the enterprise knowledge system. You'll define the canonical metric definitions, causal relationships, and operational context models that every AI and analytics product consumes. Every Concept will mean the same thing in every dashboard, every business review slide, and every AI response - because you made it so. You'll scale this across operational domains (routing, capacity, labor planning, delivery execution) and resolve the genuinely hard problem of conflicting definitions across organizations.

Own the causal attribution framework. We maintain root-cause attribution systems that decompose metric misses into specific, quantified drivers - separating planning failures from execution failures so leaders know exactly which lever to pull and who should pull it. These aren't academic models. They power daily and weekly business reviews, auto-generated executive briefings, and AI-driven action recommendations. When the attribution is wrong, senior leaders notice in real time. You'll drive accuracy, expand coverage, and build the closed-loop learning system that tracks which recommended actions delivered results.

Scale the self-service platform. We recently launched a browser-based platform that compressed the knowledge onboarding process from 6-8 weeks to under an hour - a 98% reduction in cycle time. Over 100 analysts and knowledge creators will use it to define metrics, visualize the knowledge graph, create and test new nodes, and deploy - all without setting up a local development environment. Your job is to evolve this from an internal team tool into an enterprise-grade knowledge management platform that any analytics team can use. That means governance at scale: who can define a metric, who can modify a causal relationship, how you handle conflicting definitions, and how you maintain quality as the contributor base grows from 100 to 1,000+.