Best Practices

Maximize the quality and efficiency of AI-generated test cases with these proven strategies and recommendations.

Getting the Most from Reqase Lite

While Reqase Lite automates test case generation, following these best practices ensures you get high-quality, consistent results that align with your team's testing strategy.

Choosing the Right LLM

Not all AI models are optimized for speed. Some are slow and designed for deep analysis, while others prioritize quick responses.

Recommendation

In most cases, a faster, cheaper model delivers results just as accurate and is better suited for test case generation.

Consider these factors:

Speed: Faster models reduce wait time and improve team productivity
Cost: Test case generation doesn't require the most expensive models
Accuracy: Mid-tier models often provide sufficient accuracy for structured test cases

Understanding the Three Levels of Context

When generating test cases, Reqase Lite passes information to the LLM in three layers. Use all three together for the most consistent results.

System Level

Defined within the plugin and determines the required format (BDD, Manual, or Generic).

This level is automatically configured based on your test format selection.

Test Case Rules

Ensures consistency across your team or organization (naming conventions, structure, terminology).

Tip: Define clear test case rules in your project settings to maintain consistency.

User Story Context

Provides scenario-specific details such as acceptance criteria or unique conditions.

Tip: Add additional context in the generation dialog for complex scenarios.

Writing Effective Jira Fields

The quality of generated test cases depends heavily on the clarity of your Jira issues.

Do's

✓Write acceptance criteria in plain, specific language
✓Include edge cases (invalid inputs, boundaries, error conditions)
✓Use concrete examples and specific values
✓Define clear expected behaviors and outcomes

Don'ts

✗Avoid vague terms like "System should be secure"
✗Don't use ambiguous language or unclear requirements
✗Avoid missing edge cases or error scenarios
✗Don't leave acceptance criteria incomplete

Example: Good vs. Vague

❌ Vague

"System should be secure"

✅ Specific

"System locks user account after 3 failed login attempts"

Reviewing AI-Generated Cases

AI accelerates test creation, but human review remains essential to ensure quality and alignment with your testing strategy.

Reviewer Checklist

Alignment: Do steps align with acceptance criteria?

Coverage: Are both positive and negative cases included?

Specificity: Are expected results specific and measurable?

Strategy: Does the case match the project's test strategy?

Completeness: Are all necessary preconditions and test data defined?

Supporting Exploratory Testing

Reqase Lite can help teams generate exploratory test charters, expanding its usefulness beyond scripted test cases.

How to Use for Exploratory Testing

Provide a charter template in your test case rules to standardize generation across your team
Use in conjunction with Generic or Plain Text type test cases
Generate test ideas and areas to explore based on user stories

Example Charter Template: "Explore [feature area] to discover [risk/concern], using [test approach] and focusing on [specific aspects]."

Limitations to Keep in Mind

Understanding the limitations of AI-generated test cases helps set realistic expectations and guides effective usage.

Redundancy

AI may produce redundant or obvious cases — filter and consolidate before saving.

Domain Specificity

Domain-specific details may still need manual editing to match your industry or technical context.

Not a Replacement

The plugin does not replace risk-based analysis, exploratory testing, or strategic test planning.

Key Takeaways

Choose faster, cost-effective LLMs for test case generation
Leverage all three context levels for consistent results
Write clear, specific Jira fields with concrete examples
Always review AI-generated cases before finalizing
Understand limitations and use AI as an accelerator, not a replacement