AI-Powered Risk-Based Testing: QualityKiosk Helps a Leading APAC Insurance Company Accelerate the Testing Lifecycle and Elevate Quality

Our approach relies on an advanced AI learning-based defect predictor. This machine learning marvel undergoes a meticulous training phase, absorbing insights from the data of previous projects. During the testing phase, it accurately anticipates potential defect-free and defective modules within the new project. This predictive power ensures a proactive stance in addressing issues before they arise.

We’ve harnessed the potential of AI-based data analytics to revolutionize risk-based testing. By employing sophisticated algorithms, we identify and thoroughly examine the most critical areas, creating a comprehensive heat map of business and application hotspots. This strategic insight guides our testing efforts, ensuring a laser-focused approach to quality assurance. This early detection empowers us to proactively mitigate risks, ensuring a robust and flawless project execution.

Lessons Learned

Our experience and research have shown that defect predictors are invaluable tools during new project implementations. They not only save time and money but also accelerate the achievement of business goals, surpassing competitors in terms of efficiency and effectiveness. The model has been in use at the company for twelve months now, and the prediction results given above were so satisfactory that the quality assurance team at the client decided to integrate the model into their configuration management system. They planned to use the prediction model before the testing phase so that the defect-prone files would be investigated by the developer before transferring the project to the test team.

We suggest that such predictors should be used before the testing phase to guide the testers through defect-prone modules in the software system. QA teams can continuously refine their methods. This iterative improvement process aligns with Lean QA principles, where the focus is on eliminating waste and optimizing efficiency in the QA process.

Maintenance

Similar to many AI-based models, our model also requires calibration. The company decided to train the model with new data regular interval and make predictions on new releases. Since the model has been successfully integrated with the company’s software system, the quality assurance team is
selected to form a new training set every month and update the model with new parameters. The company also motivates the teams to make such tools part of their routine during the development and testing stages. This way, it will be easier to apply the model in collaboration with the development and test teams to analyze the code quality and predict the critical parts of the software