Record, generate, run: AI-powered UI test generation for iOS
Article Summary
Grab Engineering just turned a days-long iOS testing nightmare into a 10-minute workflow. Their AI-powered system records your app interactions and auto-generates complete UI tests, including mocks, feature flags, and analytics verification.
Grab's mobile team built an AI workflow that transforms recorded user sessions into executable Swift UI tests. The system leverages their AutoTrack SDK data, combining a local proxy recorder with AI code generation to produce production-ready test files that integrate seamlessly with existing infrastructure.
Key Takeaways
- Test creation drops from days to 10-20 minutes per flow
- AI generates three linked files: test code, API mocks, and feature flags
- System captures real network calls and converts them into mock expectations
- Works with existing test infrastructure, no new tooling required
- Human review still essential: AI creates starting point, not production-ready code
Grab reduced iOS UI test creation time by over 95% while maintaining quality through a record-generate-review workflow that plugs into their existing test stack.
About This Article
Grab's iOS developers had gaps in test coverage. Teams focused on the critical user flows but skipped edge cases, which meant bugs were hard to reproduce in production without the full picture of how users actually moved through the app.
Grab built a Mobile UI Testing AI Workflow that combines a local proxy test recorder with AI-assisted code generation. The system uses custom prompts tuned to their Swift test structure, base classes, and helper utilities.
The workflow generates three linked files: test code, API mocks, and feature flags. It includes analytics verification so developers can check that expected events fire during test execution through an instrumentation server.