EMA (Enterprise Management Associates) recently released a report entitled “The Economic Collapse of Software Testing with AI”. In this report, author Torsten Volk, Managing Director of EMA, explains why traditional approaches to software quality cannot be extended to meet the needs of modern software delivery. He highlights the five major categories of AI and the six key issues of test automation that AI addresses.
We talked to Torsten about his reports and his insights into the impact of AI on software testing.
Q: What’s wrong with the current test status? Why do you need AI?
Organizations that rely on traditional test tools and methods are unable to meet the needs of today’s digital demand and are rapidly lagging behind their competitors. The complexity of applications and the increased time it takes for businesses to reach market makes it difficult for software distribution teams to keep up. There is an increasing need to use AI to optimize processes, eradicate routine and repetitive tasks, and control the cost of out-of-control quality.
Q: How and what does AI help?
There are five key features that AI can help. Smart scrolling / Natural Language Process (NLP) -led test creation, self-healing, coverage detection, anomaly detection, and visual inspection. The report I wrote highlights six key issues where these features can help. Examples: false positives, test maintenance, inefficient feedback loops, increased application complexity, chaotic device increases, toolchain complexity.
While major organizations have already adopted some degree of self-healing and AI-driven test creation, the most influential is visual inspection (or visual AI), which covers the user experience completely and accurately. You can learn and adapt to new situations without having to write and maintain code-based rules.
Q: Do people adopt AI?
Yes, the adoption of AI is increasing for a variety of reasons, but for me it’s not that people aren’t adopting AI, they are adopting technical features based on AI. For example, people want the ability to perform NLP-based test automation for specific use cases. People are interested in the ROI that comes from the speed and scalability of leveraging AI in the development process, not necessarily how sausages are made.
Q: How will the role of developers / testers change with the implementation of AI?
When looking at test automation, developers and testers need to make decisions about what belongs to test automation. For example, how are they classified? Then you basically just set up a framework for AI to work, provide feedback, and continuously improve performance over time.
When this happens, developers and testers can do more creative, interesting and valuable work by eliminating the effort of routine and repetitive tasks. This work itself is not worth it, but it must be done correctly every time.
For example, check the rendering of thousands of web pages. Some of them make little difference, but they don’t matter. If you can filter out everything that’s okay with your machine and highlight some that you don’t know if it’s a defect, you can reduce your work from thousands to very few.
Auto-classification is a good example of how you can reduce your work. If you’re reducing repetitive work, that means you don’t miss things. On the other hand, if you look at the same thing, it looks like the same page every time, so you may miss something. If you let AI tell you this page, it’s a little different from the other pages you’ve seen so far. As a result, iit eliminates repetitive and mundane tasks and reduces the likelihood of error-prone results.
Q: Do I need to hire an AI expert or develop in-house AI practices?
The simple answer is no. There are many vendor solutions that can leverage the AI, machine learning, and training data already available.
If you want to implement AI yourself, you really need people with two sets of domain knowledge. One is the domain required for AI applications, and the second is a deep understanding of the potential of AI and how to chain them together. Functions together. Often it is too expensive and too rare.
If the core artifact is not an AI artifact, but an ROI artifact that AI can provide, it’s a good idea to find a tool or service that can do it so you can focus on your domain’s expertise. increase. This makes life much easier, as there are far more people in the enterprise who understand the domain, and only a handful of people who only understand AI.
Q: You said that the visual inspection function has the greatest impact, but how can it help you?
Training deep learning models to inspect applications through the eyes of the end user is important to eliminate many common repetitive tasks that cause human inefficiencies.
Smart crawl, self-healing, anomaly detection, and coverage detection are each point solutions that help organizations reduce the risk of blind spots while reducing human workload. However, visual inspection goes a step further in order to understand the workflow and business requirements of the application.
Q: Where should I start today? Can AI be integrated into existing test automation practices?
Yes – Applitools Visual AI example.
Q: What is your future situation?
Autonomous driving is a vision for the future, but you have to ask yourself. Why don’t you have a self-driving car yet? That’s because we still connect models to each other today. But in the end, AI handles all tactical and iterative decisions, making it more valuable from a business-focused perspective for humans to think more strategically at the end of the process. I am aiming for that.
Thanks to Torsten for spending time with us and if you are interested in reading the full report http://applitools.info/sdtimes ..
Destroy the economics of software testing with AI
Source link Destroy the economics of software testing with AI