Calvin Schroder: AI Powered Unit Testing: The Rise, The Challenges, and The Future
summary
Evaluates 30+ AI test tools and chooses GitHub Copilot for companywide rollout, reporting 87% retention and ~15–30 minutes saved per 8‑hour coding day. Builds a production "unit test improver" that uses Azure OpenAI GPT‑4 with a RAG pipeline backed by a graph DB (neo4j) and LangChain style prompts: preprocessor decides improvability, generator emits improved tests, reviewer approves or retries. Works on Java, Python and some JS tests and reduces human iterations while keeping costs manageable.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 1:09AI initiatives at DT
- 2:34AI-generated unit test challenges
- 5:27RAG for AI context
- 7:14Commercial AI testing tools
- 10:50IDE plugins for AI testing
- 12:58Demo: GitHub Copilot unit tests
- 15:05Copilot rollout feedback
- 19:01Limitations of AI testing tools
- 19:22Unit Test Improver intro
- 23:37How Unit Test Improver works
- 27:06Benefits of Unit Test Improver
- 28:55Building AI testing solutions
- 29:57Recap and key takeaways from the talk
- 30:20AI beyond unit testing at DT
- 31:45Q&A
related talks
57mOne Team’s Journey to Using AI Without the Hype
41mAI-Augmented Testing: How Generative AI, Prompt Eng Turn Testers Into Superheroes
26mAI and the Future of Testing
33mEnsuring quality in an AI-driven world
57mAccelerate Your Developer Productivity with AI Embrace the Future Now!
36mUnleashing Generative AI: Improving Developer Productivity in SRE
01h 02mPANEL - Unleashing the Future: The Impact of Gen AI on Development Productivity
32mFast, Smart, and Fragile: Rethinking Testing Leadership in the Age of AI