It is important to understand that AI is not a universal solution. Within the industry, there is still work to be done to identify the use cases with the highest ROI. Having a method of implementation and governance to accompany the adoption of AI in testing is often the safest and most productive way to obtain benefits from this new technology.
So, when asked if the tester is about to disappear with the arrival of AI, the answer is a resounding no. But, as has happened before with other disruptive technologies, quality engineers will need to adapt to the use of these —and many other— new tools to remain relevant in the IT market.
These activities include:
- Adopting new tools such as public LLMs to train in prompt engineering.
- Training in context engineering.
- Incorporating new tools into daily workflows in testing activities.
- Studying new techniques for testing software that incorporates AI functions.
- Always considering regulatory issues defined in the AI Act, NIST and other bodies.
- Integrating new architectures that result from AI implementation in software, such as agentic architectures.
What is the future of testing processes and AI?
Just as the profile of quality engineers has evolved, the testing process will adapt to the technologies, processes, and frameworks that deliver the best results. In fact, we are already witnessing how the testing process is changing as AI-based technology proves to be competitive and mature enough for implementation in the testing environment.
Like the testing process, the software development cycle is also adapting quickly to market demands and the need to experience the benefits of AI implementation in the activities with the highest ROI.
It is worth remembering: this is still in a phase of evolution and discovery. Having frameworks for implementation and result measurement is highly valuable in supporting these implementations.
How AI is being used in testing: 6 current trends in Quality Engineering
1 – AI-assisted development for creating automation scripts
This increases tester productivity, allowing them to program scripts and improve code quality through more exhaustive validations, using an AI agent for peer programming.
2 – Code documentation
Generative AI has proven particularly useful in creating documentation for functions and libraries, as well as in adding comments to key pieces of code, making it easier for other test engineers working in the same framework to understand.
3 – Dynamic regression suite
AI enables the detection of testing coverage patterns, allowing identification based on changes made to a software component and suggesting which tests to run for adequate coverage.
4 – Automatic maintenance of test cases (self-healing test cases)
Machine learning algorithms are used to detect patterns of repetitive or similar changes and determine which parameters to adjust in scripts to make them work again, with minimal human intervention.
For example, Locators in Selenium become invalid when a product’s UI changes its HTML structure. When this happens, tests begin to fail and a quality engineer must debug the script, locate the faulty Locator, find the new version on the page, extract the new Locator of the missing object, and add it to the object repository. With ML techniques, this process could be executed autonomously: the change would be proposed in a new branch of the framework and stability tests would be run on the new AI-generated code. The quality engineer would then only need to analyze and approve the change.
5 – Visual recognition and UI testing
Here, AI validates interface objects visually to confirm that they appear as they should in terms of design and alignment, in addition to functionality.
6 – Low-code automation testing
AI is integrated into tools that facilitate test flow automation using a prompt and context. This accelerates automation and reduces complexity. However, it affects maintenance and test stability, so its use is more limited. It has been applied in short-lived testing scenarios, where achieving ROI with traditional automation would be difficult.
Benefits and challenges of using AI in testing
Many of the benefits of applying AI in testing arise from the speed and productivity it provides. The most notable are:
- Reduced testing cycles and faster time-to-market.
- Increased productivity and code quality.
- Ability to perform proactive testing through error prediction in software, cloud platforms, and containers.
- Well-documented code aligned with best practices with minimal effort.
However, like any technology that has not yet reached maturity, it also has challenges and limitations. When machine learning algorithms are used, successful results depend on the quality of the dataset used for training. Moreover, datasets must be large enough to avoid bias due to limitations.
Low-code automation tools, for their part, may not always produce 100% reproducible results in a new instance, whether due to lack of context in the prompt or insufficiently clear examples. And of course, using public LLMs involves costs for API keys and tokens, as well as exposing data to third-party infrastructures. Therefore, when developing solutions involving sensitive data, it is necessary to evaluate additional security layers that comply with regulations such as the General Data Protection Regulation (GDPR) or the AI Act.
By Ángel Limas, Quality Engineering Practice Head at Baufest.