Ofertas › Zillow
Zillow · Flexible / Remote · Remoto
About the team At Zillow, we believe home should be more than just a dream. The Zillow Home Loans (ZHL) Product team delivers a home financing journey that gives consumers more confidence, transparency, and ease than they can find anywhere else. We partner closely with Engineering, AI, Design, Research, and Operations to bring holistic experience innovation to market. This team is building an AI-native operating model for service delivery - agentic systems that reason about the next best action, and durable workflow infrastructure that carries those decisions out reliably. About the role The Principal Product Manager, Directed Workflows drives the vision, strategy, and execution for how ZHL's sales and service process gets encoded as directed work. AI agents reason over each customer's situation, adapt that process to the specific opportunity in front of them, and determine the next work that should happen: the action, the outcome it is meant to create, the quality bar it must meet, and the skills required to execute it. You will own that loop across engagement and loan production. You design the AI harness (context, tools, memory, orchestration, escalation, guardrails) that determines what those agents reason over and what work they produce. You own the eval infrastructure that scores those decisions against the process you defined, learns from operator feedback, and uses experiments to decide what to keep, change, or retire. You Will Get To Define and lead a multi-year product vision for Directed Workflows as an AI-first service system spanning engagement and loan production Define, with Sales and Business Operations, the sales and service process this system encodes: customer jobs, value at each stage, and how that experience earns attention and advocacy Design the AI harness that turns that process into AI agent behavior: context engineering, tool design, memory, orchestration, and guardrails, including where a model should reason, where deterministic systems should execute, and what requires a person. The work the agent produces names the action, intended outcome, quality bar, and skills required to execute Own the eval infrastructure you build or integrate: score AI agent output against the intended process, learn from operator feedback in production, and run experiments that show whether the directed work moved the outcome Sequence from high-leverage pilots to platform: prove outcomes in focused increments, retire or redesign what does not move the