Ship - Hype: Rolling Out AI at Scale to 60 Android Engineers
summary
Qonto rolls AI into a 60+ Android org by starting with a safe sandbox (Dust) for compliant LLM access, then adding a repo-aware bot (30 daily users) and Cloud Code agents that run inside Android Studio. One experiment fails: Figma→Compose via MCP hallucinates and yields ~70% usable code. One succeeds: an agent reorganizes 300 test-fixture files in 5 days vs an estimated 3 weeks and halves test build time. Team standardizes reusable prompts in git and targets repetitive boring tasks to capture ~13% cycle time savings.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 3:10Conto and AI role
- 5:21Phase one: AI playground
- 11:46Phase two: Agentic AI
- 14:56AI experiments learnings
- 18:27AI adoption patterns
- 23:46AI champions and adoption
- 27:20Measuring AI impact
- 32:57Key takeaways
- 34:42Q&A
related talks
38mOne Android Dev, Ten AI Teammates: Scaling Delivery With Claude Code
41mHow AI Will Change Mobile Development
34mMulti-coding agent workflows
44mThe Swarm Next Door: Running Parallel AI Agents on a Real Android Codeba…
27mSoftware engineering with GenAI
30mHow We Moved from Android Studio to Slack
18mDispatch from the Future: building an AI-native Company
57mOne Team’s Journey to Using AI Without the Hype