SaaS • E-commerce • Fintech
AI-Powered Test Maintenance
Keep your test suite green as your product evolves — without dedicating engineers to fix broken selectors and outdated steps.
Challenge
The hardest part of E2E testing is not writing the first test — it is keeping hundreds of tests passing as the application changes. Selector updates, copy changes, redesigned flows, and minor layout shifts break tests constantly. Teams report spending 10-20 hours per month just maintaining existing tests, and when the backlog of broken tests grows large enough, the entire suite gets abandoned. The investment in building test coverage is wasted.
Stably approach
Stably's AI auto-heal detects when a test fails due to a UI change rather than a genuine bug. It analyzes the failure, identifies the updated selector or flow, and applies the fix — either automatically or as a suggestion you review. The CLI command `stably fix` can repair failing tests in bulk by analyzing error context, screenshots, and page state. Over time, the suite maintains itself while your team focuses on building new coverage instead of patching old tests.
What changes
A button rename does not break 15 tests anymore
Your designer changes "Submit Order" to "Place Order." Auto-heal detects the text-based selector change across every test that clicks that button. It updates them all and shows you a diff: old selector → new selector. You approve or edit — no hunting through test files.
stably fix repairs failing tests in bulk from the command line
After a major UI refresh breaks 30 tests, run stably fix. It analyzes each failure — screenshots, error messages, page state — and generates fixes. You get a summary: "22 auto-fixed, 5 need review, 3 are genuine product bugs." Hours of manual work reduced to one command.
Three healing modes let you control how much autonomy the AI has
Set each test step to Strict (fail on any change), Auto-heal (fix minor UI changes like selectors), or AI (regenerate the step from its description on every run). High-risk assertions stay strict. Routine navigation steps auto-heal. You decide the threshold.
Broken tests are triaged, not just flagged
When a test fails, Stably classifies it: is this a broken selector (auto-fixable), a UI flow change (needs review), or a genuine product bug (needs an engineer)? The triage saves your team from treating every red test as a fire drill.
When this is the right fit
- Engineers spend more time fixing tests than writing new ones
- A significant portion of the test suite is disabled or skipped
- Test failures are routinely dismissed as "probably just a selector change"
- Past investments in E2E testing were abandoned due to maintenance cost