Supporting track

AI in Testing

Practical uses that actually survive a code review.

What this track covers

Test case generation from requirementsSelf-healing locatorsVisual regression with AI diffingFailure triage & clusteringTest data generationRisk-based test selectionLimitations & review discipline

AI-assisted failure triage

Cluster failures by stack-trace and error similarity so 40 red tests become 3 root causes.

Why it is required: Triage time, not authoring time, is the real bottleneck in large suites.

Example: Group failures by normalised exception signature before a human looks at them.

text

140 failures →
2  cluster A (31)  TimeoutException on /checkout  → deploy changed the pay button id
3  cluster B (6)   401 Unauthorized               → expired service account token
4  cluster C (3)   AssertionError on total        → genuine pricing defect

Expected result

Three investigations instead of forty.

Common mistakes

Trusting self-healing locators without review (hides real UI defects)
AI-generated tests merged without human assertion review
Sending production data to external models

Where would you use AI in your automation strategy, and where would you not?