Best of Both Worlds

​

Overview

​

Why Hybrid?

​

The Hybrid Workflow

​

Phase 1: AI Generation

1. Describe your test goal to the Test Planner Agent
2. Review the generated plan - don’t accept blindly
3. Identify what needs refinement - locators, assertions, timing

​

Phase 2: Manual Refinement

1. Fix locators with Element Picker
   • Replace generic selectors with stable ones
   • Add data-testid attributes where possible
2. Strengthen assertions
   • AI might add basic verifications
   • Add specific checks for your business logic
3. Handle edge cases
   • Add waits for dynamic content
   • Configure timeouts appropriately

​

Phase 3: Record Missing Pieces

1. Complex UI components - date pickers, dropdowns, drag-and-drop
2. Multi-step flows - where clicking through is faster than describing
3. Capturing exact user behavior - mimicking real user patterns

​

Phase 4: AI Assistance

1. AI Chat to add assertions you might have missed
2. Suggestions for edge cases
3. Refinement of step descriptions

​

Practical Examples

​

Example 1: E-commerce Checkout

Step 1: AI Generation
─────────────────────
Prompt: "Create checkout test with valid payment,
        invalid card, and expired card scenarios"

AI generates 3 tests with basic structure

Step 2: Manual Refinement
─────────────────────────
- Fix product selector: .product-card → [data-testid="product-1"]
- Add precise total verification
- Configure payment form waits

Step 3: Record Payment Flow
───────────────────────────
- Record the 3D Secure popup handling
- Capture exact payment form interactions

Step 4: AI Enhancement
──────────────────────
Ask AI Chat: "Add verification for order confirmation
             email reference number"

​

Example 2: User Registration

Step 1: Record the Flow
───────────────────────
- Record complete registration
- Captures form fields, validation, success

Step 2: AI Enhancement
──────────────────────
Ask AI: "Generate edge cases for this registration test"

AI suggests:
- Invalid email format
- Password too short
- Duplicate email error
- Missing required fields

Step 3: Manual Precision
────────────────────────
- Fix validation message selectors
- Add specific error text assertions
- Configure unique email with {{ random.email() }}

​

Example 3: Search Functionality

Step 1: Manual Setup
────────────────────
- Create test structure
- Add navigation to search page
- Configure environment variables

Step 2: AI Chat
───────────────
"Add steps to test search with:
 - Valid query returning results
 - Query with no results
 - Special characters in query"

Step 3: Refine Manually
───────────────────────
- Use Element Picker for result items
- Add count verification
- Configure wait for search results

​

When to Use Each Method

​

Use AI When

​

Use Manual When

​

Use Recorder When

​

Tips for Success

​

1. Don’t Trust AI Blindly

• Review generated locators
• Verify assertions match your expectations
• Test the generated steps

​

2. Let AI Do Repetitive Work

• Navigation sequences
• Form filling patterns
• Basic verification structures

​

3. Keep Manual Control Over Critical Paths

• Choose locators manually
• Write assertions explicitly
• Test thoroughly

​

4. Iterate Quickly

Generate → Test → Refine → Test → Enhance → Test

​

5. Document Your Decisions

• Note why in the test description
• Help teammates understand the choice
• Guide future AI interactions

​

Element Picker in Hybrid Workflow

​

After AI Generation

1. Run the AI-generated test
2. Find steps with unstable locators
3. Use Element Picker to fix them
4. Verify the fixed steps work

​

During Recording

1. Record captures basic locators
2. Pause recording
3. Use Element Picker for critical elements
4. Resume recording

​

Locator Priority

1. [data-testid="..."] - Most stable
2. [role="..."][name="..."] - Semantic and stable
3. Unique CSS classes - If stable in your app
4. Text content - Only if text won’t change

​

Common Patterns

​

Pattern: AI Structure, Manual Polish

1. AI generates test skeleton
2. Manual: Fix all locators
3. Manual: Add precise assertions
4. AI: Suggest additional checks
5. Manual: Final review

​

Pattern: Record, Enhance, Verify

1. Record the user flow
2. AI: "Improve this test with better assertions"
3. Manual: Review and fix suggestions
4. Run and verify

​

Pattern: Manual Critical, AI Peripheral

1. Manual: Build the critical path test
2. AI: "Generate variations of this test"
3. Manual: Review and adjust variations
4. Run all variations

​

Troubleshooting Hybrid Workflows

​

AI Suggestions Don’t Fit

• Provide more context in your prompts
• Reference specific elements by name
• Describe your application structure

​

Recorder Misses Steps

• Slow down your interactions
• Add missing steps manually
• Use AI Chat to fill gaps

​

Locators Keep Breaking

• Always verify with Element Picker
• Add stable test IDs to your application
• Use multiple fallback locators

​

Next Steps