The 6 principles of conversation-driven development
Conversation-driven development is a set of practices that help teams build more resilient, more natural-sounding virtual assistants. It’s a user-centered approach that’s baked into every feature Rasa ships.
Learn about users’ needs by sharing your assistant with test users early in development. Train your assistant on test users’ conversations to improve accuracy.
Read conversation transcripts between your assistant and your customers. Use conversation reviews to see the impact of model changes and updates.
Use both internal metrics and external signals to measure how well your assistant is helping users. Tag conversations to track successes and what’s not working.
Source training data from real-life conversations by annotating user messages. Create a virtuous cycle of learning from what users actually say.
Apply improvements and bug fixes in short, iterative development cycles. Create test cases from successful conversation flows.
Strive for complete test coverage across your assistant. When changes are pushed to production, run automated tests to catch regressions.