Zack Reneau-Wedeen is head of product at Sierra, where he develops conversational agents that help businesses serve customers, complete transactions, and operate across chat and voice. Previously, he was a founding product manager for Google Lens and Google Podcasts.
He studied computer science and economics at Yale and earned an MBA from Harvard. During seven years at Google, he worked on Search and helped build Lens, testing early image-recognition models against plants and flowers and confronting a problem that would shape his later work: AI systems can produce impressive results without performing reliably. He subsequently helped bring Google Podcasts to iOS and the web and supported a creator program developed with PRX to fund and mentor underrepresented podcasters.
After Google, Reneau-Wedeen worked in product leadership at Robinhood and CoinTracker, leading product and design at the latter, before joining Sierra. His work there focuses on turning conversational models into dependable products that can check inventory, recommend purchases, track orders, issue refunds, and manage subscriptions.
- The Agent Development Life Cycle: Reneau-Wedeen applies software-development discipline to inherently unpredictable models: monitor real conversations, inspect failures, collect customer feedback, convert problems into regression tests, and release improvements deliberately. His framework for shipping and scaling agents combines flexible model behavior with deterministic software controls; his AI Engineer Summit talk illustrates the approach through a retail agent that handles sizing, inventory, order tracking, and refunds.
- One agent per brand: He favors a single customer-facing agent that retains conversational context and customer history across support, sales, and transactions, while coordinating specialized models and tools behind the scenes. Dividing those interactions among separate agents risks reproducing a company’s organizational silos in the customer experience. His approach to production agent architecture also connects commercial models to completed customer outcomes.
- Voice as a responsive interface: He treats voice and chat like responsive versions of the same product: shared business knowledge and behavioral policies, adapted for each channel’s timing, phrasing, and interaction patterns. His work on more human voice experiences emphasizes latency, interruptions, turn-taking, and anticipating callers’ needs, with voice-to-voice models offering richer interactions than transcription alone.
- Agent Studio 2.0: He helped shape Sierra’s collaborative agent-development platform, whose Journeys, Workspaces, and Integration Library let engineers and customer-experience teams build sophisticated workflows, coordinate changes, and connect agents to business systems.
