As of October 2025, Yelp’s AI PM workflow, as demonstrated by Priya Badger, shows how defining 'golden conversations' in Claude can be a critical step in creating high-quality prototypes for user interactions. By outlining ideal user-AI dialogues upfront, PMs can ensure that their product design aligns closely with customer needs before moving into full development.
Here’s a step-by-step guide on how to utilize this workflow effectively: 1. Draft Golden Conversations: Start by using Claude to draft sample dialogues (or 'golden conversations') that represent the ideal interactions between users and the AI. Focus on specifics such as handling image uploads, recognizing visual issues (like a cracked porch), and prompting users for details. Document these interactions clearly. 2. Transition to Prototyping: Leverage Claude’s artifact generator to convert these golden conversations into a working prototype. This includes embedding system prompts and making live calls to the language model, allowing you to simulate actual user interactions. This step helps in creating a proof-of-concept without the immediate need for API integrations. 3. Refine the Interface: Use Magic Patterns’ inspiration mode to quickly generate and iterate on multiple UI variations for features such as the photo-upload function. This natural-language based prompt mechanism significantly accelerates the design phase, ensuring that the visuals and interactions meet your team’s expectations. 4. Collaborative Testing and Feedback: Once the prototype is created, set up sessions to test the chat interactions and image analysis components with your internal team. Gather actionable feedback and iterate on the design until it meets the desired user experience standards.
This workflow not only speeds up the initial phases of prototype development but also helps in validating key user interactions before full-scale implementation. While detailed performance metrics are still emerging, early implementations suggest that this approach reduces the time needed to develop interactive prototypes substantially, potentially cutting down iterative cycles by several days. AI PMs adopting this method are better positioned to maintain a high level of quality in design and user experience.