Hilary Gridley shows how to find overlap between features to build smarter user journeys, especially when you have a new AI feature with low adoption that executives want people using. The approach: Use an LLM to understand user characteristics across features. At Whoop, she looks at demographic and behavioral information. For B2B, you’d examine industries, job titles, or seniority. The goal is discovering which features have the strongest user overlap with the one you’re trying to promote, then generating persona-based journeys between those features. Here’s Hilary’s workflow: First, pull chaotic data sets from Amplitude - different features with user breakdowns by age, gender, and other attributes. Paste everything into a Google Sheet. It won’t be clean for traditional analysis, but that’s fine. Second, load this messy data into Claude. In her example, she has an AI meeting assistant plus features like task manager, document collaboration, and pulse check with user attribute breakdowns across them. Third, prompt Claude to analyze the user mix across features and identify which has the most in common with the new AI feature. The analysis reveals patterns like “task manager has highest overall similarity with AI meeting assistant” and “managers show highest engagement with AI meeting across all teams.” Fourth - and this is key - take it to the next step. Hilary always asks: “Great analysis. What should we do about it?” Ask Claude to build three extremely specific user journeys personalized to the personas you’ve identified. You’ll get different ways to connect people from one feature to another, complete with example message copy and CTAs you can start testing right away. ➡️ Skip the clean data requirements. LLMs can find patterns in messy, real-world data exports and turn them into specific, testable user journeys between features. Check out Hilary’s course (not sponsored, she’s just awesome).