Reinventing App Testing with Agentic Apps
summary
Sophie demonstrates agentic app testing that runs ~1,000,000 tests/month and uses LLMs plus UI XML/screenshots to generate, execute and triage tests on real devices in Seattle and Dallas. It adapts nondeterministic flows, auto-handles permission popups, enforces step performance thresholds (≈1,000 ms), and claims to cut the 2–3 hours/day manual failure investigation. The system also authorizes bidirectional export and coding — converts agent tests to Appium/Python/Java and produces accessibility checks and failure stack-trace analysis.
AI-written summary · about
chapters (10)
- 0:00Agentic app testing intro
- 1:28LLMs in mobile test automation
- 3:14Live demo: agent-driven tests
- 7:27Adapting tests with specs
- 9:12Converting tests to scripts
- 12:03Unifying no-code and agentic
- 13:29No-code test automation
- 15:16Performance thresholds
- 16:41Test results and agent training
- 17:23Future of agentic testing & Q&A
related talks
47mAgentic Testing: Your Skilled Human Tester
19mThe GPT-ification of Mobile App Testing
28mTesting at the Speed of Thought: QA in the Era of GenAI apps
31mHow Agentic AI Is Reshaping Mobile App Quality
23mWhy testing is also a developer's problem
31mKeynote: How Software Testing Can Increase Agent Autonomy
32mWhy testing is a developer's problem
43mTesting Agentic AI Applications: Beyond Traditional QA