Introducing Agent Testing
summary
TestMu AI shows a beta agent-to-agent testing platform that ingests PRDs, Confluence and Jira, auto-generates hundreds of scenarios (intent recognition, conversational flow, multimodal) via parallel agents, and evaluates a target bot on probabilistic metrics like hallucination rate and conversational completeness. Tests run on HyperExecute, keep a human-in-the-loop for prompt/data tuning, and surface concrete failures (repeatedly asking for an order number, fabricating policies) so teams can iterate before production.
AI-written summary · about
chapters (5)
- 0:00Welcome
- 3:03Why testing agents matters
- 11:55LambdaTest agent testing
- 22:35Evaluation and results
- 25:23Q&A
related talks
43mTesting Agentic AI Applications: Beyond Traditional QA
37mAI agents testing How to evaluate the unpredictable
23mUnit testing LLM Agents
46mAI Testing Beyond the Basics: Ensuring Truthful and Reliable Agents
14mWelcome to TestMu Conference 2026
21mEffortless Game Testing With Planning Agents
32mQuality engineering in the AI age: From testing AI to leveraging GenAI
25mWho Let the Bots Out? A Guide to Evaluating AI Agents
