Case Study: Serve AI boosts containment and ships a better voice agent with LiveKit Agent Simulations

A LiveKit Case Study

Preview of the Serve AI Case Study

Serve AI boosts containment 15% and ships in one week with LiveKit

Serve AI, an AI provider for field service companies, faced a challenge when migrating their complex voice agent from a graph-based architecture on Retell to a single-prompt system on LiveKit Agents. The hard part was ensuring the new agent performed at least as well as the old one before switching their customers over, which required a robust way to test and evaluate its behavior.

Using LiveKit Agent Simulations, Serve AI was able to test the new agent against a hundred real production call scenarios, iterating until it reached a 98% pass rate. This gave them the confidence to ship the new LiveKit-based agent, which not only achieved parity but increased call containment by 15%, saving a specific customer an estimated $320,000 annually.


View this case study…

LiveKit

7 Case Studies