Move past a simple eval score to get a rigorous, holistic picture of how agentic features perform and improve.
Version prompts, models, and settings with AI configs.
Compare agent versions with offline evals.
Learn from real customer interactions with online evals.
Run experiments to measure impact on task completion, conversion, retention, and more.
When eval scores go up, does that actually make your product better? The only way to know whether a change to an AI feature truly works is to validate it with real users and real outcomes. Agent Experimentation lets you connect what scores well with what works for your customers, and use what you learn to improve the next version of your agentic feature.
Agent Experimentation is now available to everyone in Statsig. Learn more in the Agent Experimentation docs or the launch blog.
Keep an eye on your inbox for the latest Statsig product updates.