As software development becomes more agile and constant, classical methods of QA control are no longer sufficient. The traditional approach, based on testing software at the end of the process, can no longer keep up with the speed, complexity and high expectations of modern systems.
Welcome to the renaissance of quality control.
At Exceltic, we believe that testing is not just a final checkpoint, but a continuous and intelligent process driven by AI, data and resilience thinking. As Diogo Gonçalves Candeias presented at EXPOQA25, this new era of Quality Engineering (QE) is transforming the way organisations build, test and operate software.
From Quality Assurance to Quality Engineering: A Paradigm Shift
Traditional quality control has always been reactive and focused on detecting errors at the end of the cycle. And what does this mean? Higher costs, limited test coverage and a constant struggle to detect errors in time.
Unlike the traditional approach, quality engineering is proactive and incorporates shift-left and shift-right practices, testing at the beginning and end of the lifecycle. It also adopts tools such as AI-driven observability and chaos engineering to validate system behaviour under real-world conditions.
Instead of looking for bugs at the end, quality engineering focuses on preventing defects, implementing quality metrics throughout the software lifecycle and ensuring that testing is continuously integrated. This proactive approach raises the quality of the final product and accelerates the delivery of reliable software.
Generative AI in Testing: Automating What Matters
One of the most promising developments in QA is the integration of Generative AI into software testing. With tools such as ChatGPT, Copilot and custom LLM, teams can:
- Automatically generate test cases from user stories
- Write robust test scripts using tools such as Playwright, Cypress or Appium.
- Detect and classify errors by reviewing logs and code.
- When application changes occur, the AI can automatically adjust or rewrite test scripts, reducing maintenance effort.
This shift allows QA teams to focus on strategy and critical thinking, while AI takes care of repetitive and error-prone tasks.
And that's not all. With predictive modelling, teams can anticipate faulty tests, rerun only stable ones, and even roll back deployments if performance declines: enter the era of self-repairing infrastructure in QA.
Agentic AI: application of autonomous AI agents to software testing
Unlike traditional automation tools, which follow predefined scripts, these AI-powered agents can make decisions on their own, exploring the application under test in a more intelligent and adaptive way. In essence, they are AI systems designed to operate with minimal human intervention: they sense the application environment, reason about what actions to take, and then execute them to achieve specific test objectives.
In the context of QA, agentic AI manifests itself as a "virtual tester" capable of performing complex tasks autonomously. Some notable capabilities of these agents include:
- Automated exploratory tests: no need for a pre-established script.
- Intelligent test case prioritisation: Based on data (such as most critical areas of the system or failure histories), AI prioritises tests with the highest probability of finding errors or with the greatest impact on the user.
- Continuous learning: The agent learns from the results of each test run. If it discovers a failure, it adjusts its strategy to focus on similar areas in the future.
- Proactive approach to testing: Instead of waiting for failures to occur, an autonomous agent actively looks for unusual conditions or edge cases.
Thanks to agentic AI, the QA team gains a tireless ally that expands test coverage and accelerates problem detection. An AI agent can execute thousands of test interactions in hours that would take a human days, and do so by adapting on the fly.
Synthetic test data: Private, scalable and more intelligent
Many organisations continue to test only 20-30 % of their code before releasing it to the market. The reason? Because it is difficult to produce realistic test data. In QA, representative test data is vital to simulate real-world scenarios. However, using production data poses privacy and compliance risks.
Synthetic test data for quality control change this situation.
By using techniques such as CTAB-GAN, VAE or diffusion modelling, teams can generate high-fidelity datasets that:
- Respect data privacy (GDPR, HIPAA, etc.)
- Cover rare extreme cases
- They are cheaper and safer than production data.
- Different user profiles or specific transactions can be modelled, making testing more realistic.
Anyone, not just data experts, can now provide compliant, on-demand test data sets, improving coverage and speeding up delivery.
AI-enhanced observability
A fundamental change in the QA approach is the adoption of observability as an integral part of the quality process. Teams can continuously learn from the actual behaviour of their applications, adjust their tests based on real usage data and ensure that the system meets expected service levels under real-world conditions.
Advanced observability implies going beyond basic monitoring:
- Detecting anomalies before users are aware of them
- Predict failures (e.g. disk space, latency spikes).
- Reducing alert noise by correlating related events
- Determine root causes through NLP and chart analysis.
In short, integrating observability into QA transforms quality management into a more proactive rather than reactive process.
Intelligent chaos engineering: Breaking things with a purpose
What if your software could prepare for disaster?
Intelligent chaos engineering involves injecting faults in a controlled manner into a system to observe how it responds, and using AI to do so intelligently. Instead of manually designing chaos tests, AI analyses the system architecture, identifies weaknesses and automates fault injection scenarios.
This leads to:
- Continuous endurance tests
- Improved system design (self-healing, recovery mechanisms)
- Increased confidence in the stability of production
Tools like Chaos Monkey, Litmus and Chaos Kong are just the beginning. In modern quality control, if it can be broken, we break it first, on our terms. With each chaos experiment, the team learns and improves the architecture, achieving more resilient systems and more robust designs.
What this means for quality assurance practitioners
Far from replacing evaluators, AI is empowering them.
The future QA engineer will need hybrid skills: understanding AI models, observability tuning and orchestration of chaos experiments. Routine test maintenance? The 80% will be automated by 2026.
This evolution transforms quality control from a bottleneck to a strategic differentiator, one that improves speed, stability and customer confidence.
Conclusion: Quality is now an AI-driven advantage
The QA renaissance is here and it is reshaping the very DNA of software delivery. By embracing generative AI, synthetic data, intelligent observability and chaos engineering, organisations can achieve quality at high speed, without compromise.
In a highly competitive market, the ability to deliver high quality software in an agile manner becomes a strategic advantage. At Exceltic, we are proud to lead this transformation. If you are ready to drive your QA strategy and ensure the quality of your software, our team is ready to work together.
Are you interested in applying these practices in your software quality strategy?
Access the full presentation by Diogo Gonçalves, Head of QA at Exceltic during ExpoQA.
