Lviv AI Day 2026

More about resentation:
1. The Shift: Developers vs QA in the Age of AI

 - Developers already use AI to generate code, speeding up delivery dramatically

 - The gap: QA processes often remain manual or semi-automated

 - Risk: Testing becomes the bottleneck instead of the safety net

 - Key question: Can QA evolve fast enough to keep up?

2. Where AI Actually Helps in Testing (Not Just Hype)

 - Test case generation from:

   -- Requirements

   -- User stories

   -- Existing manual tests (e.g. Azure DevOps, CSV)

 - Automated test generation:

   -- Playwright / Selenium scripts

   -- API test generation

 - Test data generation and edge-case discovery

 - Test analysis:

   -- Coverage gaps

   -- Duplicate or useless tests

3. Where AI Can Make Things Worse

 - Hallucinated tests (invalid selectors, fake APIs)

 - Over-reuse of patterns → fragile tests

 - Misunderstanding business logic

 - Blind trust in "generated = correct"

 - AI increases volume of tests faster than quality

4. How to Evaluate If AI Works in Your Context

 - Input quality:

   -- Clear and structured requirements

   -- Consistent manual test cases

 - System characteristics:

   -- Stable UI vs frequently changing UI

   -- Well-documented APIs vs undocumented systems

 - Available context:

   -- Page Objects

   -- API documentation

   -- Existing automation framework

 - Rule of thumb:

   -- AI works best with strong context and constraints

5. Can You Trust AI With Testing

 - No you cannot blindly trust AI output

 - Treat AI as a junior test engineer working at scale

 - Validation strategies:

   -- Deterministic checks (selectors, API responses)

   -- Static validation of generated tests

   -- Cross-check with requirements

   -- Execution in controlled environments

 - Trust comes from verification, not generation

6. Architecture: How to Use AI in Testing (Production Approach)

 - Input:

   -- Manual test cases

   -- Requirements

   -- Context (repo, Page Objects, API docs)

 - AI generation:

   -- Create draft automated tests

 - Validation layer:

   -- Syntax validation

   -- Selector verification

   -- Business rule validation

 - Execution:

   -- Run tests in CI (Playwright, etc.)

 - Feedback loop:

   -- Improve future generations based on failures

 - AI must be part of a controlled pipeline, not used ad-hoc

7. Risks and Anti-Patterns

 - Letting AI generate everything without control

 - No validation layer

 - Garbage input → garbage output

 - Overconfidence in AI correctness

 - Scaling incorrect tests = scaling problems

8. Practical First Steps for QA Teams

 - Start small (one feature or module)

 - Use AI as assistant, not replacement

 - Introduce validation rules early

 - Measure:

   -- Time saved

   -- Test stability

   -- Defect detection rate

9. The Future Role of QA Engineers

 - Shift from manual test writing

 - Move to:

   -- Designing test generation systems

   -- Defining validation rules

   -- Controlling AI behavior

   -- Ensuring reliability at scale

 - QA evolves into Test Architect + AI Controller

10. Key Takeaways

 - AI is already transforming development — QA must adapt

 - AI excels in generation, scaling, and analysis

 - New risks require strict control and validation

 - Trust comes from validation, not from generated output

 - Goal: augment QA, not replace it