Rhiannon from the Finn team demonstrates how to train Finn, an AI customer service tool, to handle complex customer queries that appear simple but require multiple steps to resolve. Using a damaged order example, she shows how Finn can pull order details, ask clarifying questions, follow policies, and create solutions like return labels and replacement orders. The presentation covers the Finn Flywheel – a four-stage process of Train, Test, Deploy, and Analyse that continuously improves Finn’s performance. The training focuses on three key areas: content knowledge (what Finn knows), guidance (how Finn behaves and communicates), and procedures (structured workflows for complex multi-step resolutions).