As organizations race to implement AI in customer support, many are unintentionally testing new systems in front of the people they can least afford to disappoint: their customers. While AI can accelerate efficiency and scale, moving too quickly into production can create risk around accuracy, trust, and customer experience. The challenge is no longer simply deploying AI, but introducing it in a way that protects both customers and the business. In this interactive keynote, Assaf Barnir, Head of Global Support at Sentry, will share an engineering-driven approach to evaluating AI safely through a “Shadow Run” model. Discover how organizations can test AI alongside human teams, identify operational blind spots before they become customer issues, and establish the right signals for moving from experimentation to production with confidence. What this session will cover: - Why AI initiatives often struggle when experimentation happens directly with customers - How a “Shadow Run” model can reveal operational risks before rollout - The signals and success metrics that indicate production readiness - Turning AI evaluation into stronger feedback loops across Support, Product, and GTM teams Key takeaway: Attendees will leave with a practical framework for validating AI safely, reducing implementation risk, and moving from experimentation to production with greater confidence.