Hostinger is a leading web hosting provider serving over a million creators from around the world. The company has been working on exciting AI applications, such as an AI Customer Support chatbot that handles over 40% of support conversations. To evaluate their AI applications, Hostinger's data scientist Liucija follows three main goals: evaluating offline changes during development, assessing performance in live applications, and identifying emerging issues to improve the application. The team runs various evaluations, including component checks, automated comparisons, sentiment analysis, human review, and live user feedback. Braintrust is a tool that supports Hostinger's evaluation process, saving them hours of manual work and enabling them to develop 3x more AI features with high-quality assurance.