Artificial Intelligence, Machine Learning, and Natural Language Processing are not concepts anymore. We have been leveraging AI in testing for quite a while now. An AI-powered continuous testing platform act as an enabler to quality assurance to test faster and smarter. But what are the exact areas where AI/ML can add value to the business? The answer is, that AI can help Quality Engineering in each stage of the testing lifecycle. Listed below are a few important stages of testing.
- PointZero Automation: During Greenfield development, we typically do not have any application ready, but there are wireframe diagrams available to build the application. These wireframes are nothing but requirements for designing the application. With the latest technology like computer vision, we can automate the process of test authoring even at this stage where computer vision can be used to convert the wireframes into equivalent testing elements for every page. These test elements then can form components into a sequence of steps forming a test scenario that will further enable and simplify automation once the application is ready. This way we can have 30% – 40% of the skeleton ready for automation. Once the application is available, we just need to map these components with web elements via locators and the automation scripts are ready. This is done at the deep shift left when the requirements are on paper.
- Shift right and left: For Brownfield development, when the application is ready and we need the automation scripts to be developed in an automated way, what works best are web crawlers in a combination of NLP and AI for self-heal.
- Web crawler is a tool that points at the web application, and it then automatically crawls the entire application thus generating the user journeys by establishing links between multiple pages. While it builds the test scenarios, it also detects broken links between the pages. When the tool runs, it observes the gaps from the patterns it already learned and flags all the potential issues. This process is also called Spidering AI.
- With continuous changes happening in the application, it takes huge efforts in maintaining the automation scripts. How can AI help here? ML flags the changes in the application and notifies the tester of the change. ML can also automatically adjust to the changes. While ML does the heavy lifting, human intelligence with the right domain knowledge is necessary to do the validation of the change.
- Artificial Data Generation: While doing the causal analysis on defects, there has been a very important observation. A common cause of defects is ‘Data Unavailable’ or ‘Data insufficient’. This unavailability of data often leads to huge defect leakage and the subsequent cost to quality. How do we resolve this? This can be achieved via Artificial Intelligence. AI can generate large volumes of data needed for different business scenarios using advanced neural networks and can help with data profiling, sub-setting, masking, and injection of the data back into the test systems. This data is scalable and it is available in text, tabular, or image format, and also compliant with different data privacy laws.
- Cognitive Quality: Over a period, when the test execution happens repeatedly, a huge amount of data gets accumulated from multiple iterations of testing. There is a lot of hidden information, and many insights can be captured if the data analysis is done in the right manner and the right inferences can be hence drawn. A cognitive QA solution will bring those descriptive, prescriptive, and predictive insights from the data like defect density, defect leakage, test efficiency, risk, and quality of the portfolio and many more such descriptive KPIs can be highlighted. With an increase in continuous development and integration, there is a pressing need for continuous testing. Will it not help if we know exactly which tests to execute, what to test and automate when to stop testing, establish traceability and provide more such prescriptions? Cognitive solutions have AI/ML algorithms that can analyze the current test coverage and flag the areas having little coverage. It points to the critical business processes and/or highlights the applications that are at risk. Cognitive learning is essential and proven to improve the quality of testing that should be leveraged by organizations.
- Chatbots for quality reporting: With their 24*7 availability, chatbots are getting quite popular these days. Status reporting is the most important attribute of a successful project delivery because that highlights the issues in advance and decision making becomes effective. But it comes with its own set of challenges – no standard approach to reporting, incorrect data, misleading data, manipulated data, etc. With the help of chatbots, we can deliver the right KPIs to business users without really relying on anyone. The KPIs could be at the project level or portfolio level. Chatbots are connected to live systems, thus providing information right from the source and eliminating person dependency. With a single button click, all the important KPIs are presented by the bot and sent to your inbox on demand. This enables the business in the right decision-making.
- Causal Analysis for test prioritization: Several tickets are logged in the ticketing systems by users almost daily. These tickets result in huge data from production systems that can either be classified or clustered to draw certain patterns. Various anomalies can be identified in the data which can be consumed by Machine Learning models so that it alerts users of the underlying problems in the systems. This information further strengthens the use case for optimizing the testing and doing what is enough.
- Prod Analytics for Test Bed: There is enough information available in the application logs of any system. This storehouse of data can help us identify those important journeys executed by users a greater number of times using AI. This information will help to create the specific test scenarios and translate them into automation scripts on top of tools like Selenium or UFT etc. All this is done in a zero-touch mode leveraging the potential of AI.
- Sentiment Analysis: The sentiments of users are very important inputs to improve the quality of testing while also performing the right testing. Once we identify the sentiment patterns and the areas where users have shown concerns or negative sentiments on social media of any ticketing system, we can use pattern mining, an AI technique to perform test case prioritization and thus test those test scenarios on the critical path and on priority.
With AI-assisted test cases generation, test script generation, self-healing tests, and cognitive reporting, innovation becomes faster. In addition, it also helps organizations to achieve quality goals quickly and efficiently.
About Chaitali Lambat
Chaitali is leading the Innovation tower in Sogeti Testing Practice. She is the product owner and architect for the Testing Sales platform 'Jennie'. This is a platform that integrates assets from different areas of testing like test processes, automation, non functional along with assessment & knowledge management framework. This platform is built using PowerApps, Power Automate, MS Flow and Sharepoint. Chaitali was instrumental in architecting Business intelligence and cognitive solution called 'Cognitive QA' using Microsoft Services and tools like Power BI, Text Analytics and various cognitive services of Azure. The Cognitve QA solution has a skype bot 'Navya' which is built on top of QnA maker and LUIS that answers to testing specific queries for any account. In the past Chaitali had developed assets for process automation and also conducted several Test Process Improvement assessments using TPI Next framework. She has trained several associates on TMap test design techniques. Her whitepaper 'Aligning metrics to Org BU goals' was published in QA&Test Org in 2015 and 'Scriptless Automation - A guide to automate' was published in same forum in the year 2018..
More on Chaitali Lambat.