DEVOPSdigest

Development velocity is a standard KPI in software development, while QA velocity is rarely measured. Even with regulatory and compliance demands, DevOps methodologies, and the shifting left of numerous software testing functions, quality assurance topics (like performance, usability, security, and dependability) continue to be an afterthought in 2024.

Cloud-native technologies, distributed and serverless architectures, and low-code and no-code tools (not to mention AI/ML—we'll get to that later) have transformed software development. Testing, however, has remained roughly the same, becoming the unwanted bottleneck holding back innovation from reaching end users and limiting development velocity.

In 2024, nearly a third(link is external) of development teams cite lack of time as their most pressing challenge in ensuring code quality, which suggests that their time is dedicated to what is perceived as higher-priority tasks.

Crowdstrike and the Case for QA Velocity

The importance of software development quality assurance came into the spotlight recently when Crowstrike hit the headlines. In the preliminary PIR(link is external) published by the company, the blame was placed on a bug in the Content Validator that, upon encountering an edge case in the InterProcessCommunication (IPC) Template Type, caused millions of computers worldwide to crash.

Crowdstrike noted the list of prevention steps it will implement to mitigate the devastating fallout from a missed bug. The first thing on their list is improving the QA of software releases with local developer testing and content interface testing — steps they may or may not have included in their CI/CD pipeline in the past.

Achieving software resilience and, ultimately, software quality demands that we equip developers and testers with more advanced tools and align QA velocity with development velocity. All while remembering Kernighan's Law(link is external) and the idea that debugging is twice as hard as writing a program in the first place.

Is AI Ready to Help?

If we look at the software test cycle from an operational perspective, several time-consuming tasks are involved, with manual testing accounting for over a third(link is external) of the time spent on software QA tasks.

AI comes to the rescue with its ability to generate full test suites with unit tests and synthetic data, consolidate similar or identical tests, and potentially more. So it's no wonder that as many as 78% of software testers admit to having already adopted some form of AI tooling for tasks like test case generation and test failure analytics. AI tools can contribute to your software development QA velocity in many ways.

Intelligent Test Case Generation

By leveraging NLP and ML models, AI can discern an application's or a feature's intended functionality from its requirement documents. The model can generate diverse testing scenarios by adding historical test data, user stories, and other relevant data.

Test Data Generation

Test cases are not enough. You will need test data like boundary values, edge cases, and randomized data sets that don't accidentally use real-world data protected by data privacy regulations.

Regression Testing Automation

AI can enhance and speed up regression testing by continuously identifying code changes and automatically adjusting and evolving the test suite. It can also help prioritize the most relevant tests according to the code changes made.

Bug Detection and Defect Prediction

When fed a massive amount of historical testing data and testing metrics, AI can potentially predict defects in the software where the circumstances allow them to materialize. Moreover, AI is exceptionally good at detecting patterns. So, when provided with reliable patterns for potential bugs, AI can detect them much faster than a human tester.

User Behavior Simulation

With sufficient user behavior data, an AI can mimic real-world user behavior and generate realistic interactions with your application. In addition, AI can mimic a malicious actor just as well as a legitimate user and apply known adversarial techniques to stress-test your code(link is external) for security gaps and flaws.

That said, AI-enabled software testing is still in its infancy, and there are many limitations to its ability to offload manual work from developers and software testers in the real world. This is evident in recent surveys, where nearly half(link is external) of the survey respondents cited the lack of capable AI tools as the main obstacle to AI adoption in QA activities, alongside security and privacy concerns and the unreliability of AI tools. Even so, the same survey states that 58% of managers and senior management picked integration of AI into QA processes as the key goal in the coming years.

Between Collaboration and Autonomous Testing: What's Next for AI in Software Development QA?

The future of AI in software development QA depends on how efficiently and effectively organizations implement and integrate it into their testing lifecycles. One thing, however, is clear — AI is a robust and flexible tool to augment the antiquated processes and methods of software quality assurance and can be paired with human expertise, creativity, and contextual insight.