This content originally appeared on DEV Community and was authored by TestFort

AI in QA is a topic you can cover once a week and still miss some novelties. A year ago, we released an article on what ChatGPT can do for software test automation, and it seemed like a big deal.

Now, AI for software testing is a separate business, tech, and expert territory, with serious players, loud failures, and, all in all, big promise.

Let’s talk about

• Latest stats of AI QA testing (not everything is so bright there, by the way);
• How AI is already used to improve software quality;
• How AI and machine learning will/may be used to optimize testing;
• How to use the power of AI in testing software and reduce risks along the way.

Just to take it off the chest, we have just partnered with Viruoso AI, a top-of-the-game company that uses AI automation testing tools. It means two things:

• We are excited enough to mention it about 10 times in one article;
• We write about AI testing services from experience. We use AI to automate tests, we incorporate AI in planning manual test roadmaps, and we know how exactly tools can help software testers in the upcoming 12-18 months. We don’t plan further; this AI market advancement is crazy.

Where Do We Stand with AI Software Testing in 2024

Numbers are forgettable and often boring without context.

But they help you to see the trends, especially when they come from reliable sources. When ISTBQ, Gartner, or British Department for Science, Innovation and Technology (DSIT) cover the impact and the future of software testing with AI — you take notice.

So we give just a few numbers summarized from few research results and surveys to help you realise one thing — traditional software testing industry is living its last years.

Industry Insights and Statistics

• AI-driven testing can increase test coverage by up to 85%; Organizations using AI-driven testing reported 30% reduction in testing costs and 25% increase in testing efficiency;
• By 2025, 50% of all new software development projects will include AI-powered testing tools;
• 47% of current AI users had no specific cyber security practices in place for AI (not everything is so shiny, right?).

Real-World Proved Benefits of AI Testing
Just a brief case to show how artificial intelligence testing tools can help at any stage of QA process. They are not required for testing, true. But maybe they already should be.

We worked with a company offering to create consoles for their clients. User interface testing is paramount for such companies, but not only that. When we entered the project, we realized there were problems with bug triage, test coverage, bug report creation, requirements testing, and report creation. Using AI in software testing was new to us, but we decided to try and never regretted it. Check the numbers.

Bug Triage

• Problem. Duplicated issues and inefficiencies in assigning bugs due to multiple authors logging defects.
• Solution. Implemented DeepTriage to automate and streamline the bug triage process.
• Results. 80% decrease in analysis time and bug report creation.

Test Coverage

• Problem. Limited documentation time, predominantly covering only positive scenarios.
• Solution. Used ChatGPT to generate comprehensive test cases from requirements, ensuring better coverage.
• Results. 80% faster test case creation and a 40% increase in edge case coverage.

Bug Reports Creation

• Problem. Customer feedback needed conversion into a formal bug report format.
• Solution. Used ChatGPT to analyze and structure customer reviews into detailed bug reports.
• Results. 90% reduction in detectability and improved communication of issues.

Requirements Testing

• Problem. Need for structured user stories and consistent software requirements.
• Solution. Applied ChatGPT and Grammarly to analyze, restructure, and ensure consistency in software requirements.
• Results. 500% reduction in requirement testing time and a 50% increase in spelling mistake corrections.

Report Creation

• Problem. Time-consuming data integration from various sources during regression testing.
• Solution. Utilized Microsoft Power BI for efficient data integration and AI-driven insights.
• Results. 30% improvement in data representation and a 50% reduction in report creation time.

Our experience in implementing AI in software testing have skyrocketed since then, but it was a great start that allows us to truly believe in benefits of using AI in small and eneterprise-level projects.

https://testfort.com/wp-content/uploads/2023/04/2-AI-in-Software-Testing.png

How AI Can Be Used to Improve Software Testing

Even the best manual testers are limited by time and scope. AI is changing that. With machine learning and predictive analytics, AI enhances traditional manual testing processes. From test planning to execution, AI-driven tools bring precision and efficiency, making manual testing smarter and more effective.

Importantly, AI doesn’t eliminate the need for human testers; it helps them work more efficiently and focus on complex issues.

Test Planning and Design

Test case generation allows to analyze historical data and user stories to generate comprehensive test cases. AI is used to increase the overall coverage of the testing process (yes, large number of tests doesn’t necessarily means quality, but we still rely on human intelligence to filter trash out).

Risk-based testing relies on machine learning algorithms to prioritize test cases based on potential risk and impact.

Defect prediction is based on using AI and ML predictive models to identify areas of the application most likely to contain defects.

Test Execution and Management

Test data management will be easier with automating the creation and maintenance of test data sets using AI-driven tools.

Test environment optimization uses AI systems to manage and optimize test environments, ensuring they are representative of production.

Visual Testing is all about employing AI-powered visual validation tools (like Vision AI) to detect UI anomalies that human testers might miss.

Collaboration and Reporting

AI-powered reporting allows generation of detailed and actionable test reports with insights and recommendations using natural language processing.

Collaboration tools cover integrating AI with collaborative tools to streamline communication between testers, developers, and other stakeholders.

And now, to the most exciting part. End-to-end automated testing done right with AI-based test automation tools. It’s a mouthful, but it is exactly what you need to be thinking about it 2024.

Artificial Intelligence in Software Test Automation

Integrating AI into software testing helps get the most from automation testing frameworks. Right now, there is hardly an automated test scenario that cannot be somehow enhanced with tools for AI QA.

Self-Healing Scripts

Self-healing scripts use AI algorithms to automatically detect and adapt to changes in the application under test, reducing the need for manual script maintenance.

Dynamic element handling allows AI to recognize UI elements even if their attributes change, ensuring tests continue to run smoothly. As UI testing becomes essential to any minor and major launch, AI can assist immensely.

Intelligent Test Case Prioritization

Risk-based prioritization relies on AI to analyze code changes, recent defects, and user behavior to dynamically prioritize test cases.
Optimized testing ensures critical paths are tested first, improving overall test efficiency.

AI-Driven Regression Testing

Automated selection uses AI tools to automatically select relevant regression test cases based on code changes and historical test results.
Efficient execution speeds up the regression testing process, allowing for faster feedback and quicker releases.

Continuous Integration and Continuous Delivery (CI/CD)

Automated code analysis employs AI tools to perform static and dynamic code analysis, identifying potential issues early in the development cycle.

AI-powered deployment verification involves using AI to verify deployments by automatically executing relevant test cases and analyzing results.

Performance testing leverages AI to simulate user behavior and load conditions, identifying performance bottlenecks and scalability issues.

AI in Test Maintenance and Evolution

Adaptive test case generation uses AI to continuously generate and evolve test cases based on application usage data and user feedback.

Predictive maintenance applies machine learning to predict and address test script failures before they impact the CI/CD pipeline.

Automated test refactoring utilizes AI to refactor test scripts, ensuring they remain effective and efficient as the application evolves.

Continuous Testing

Seamless integration ensures AI integrates with CI/CD pipelines, enabling continuous testing and faster feedback.

Real-time insights provided by AI offer immediate feedback on testing results, helping teams make informed decisions quickly.

By incorporating AI into automated testing, teams can achieve higher efficiency, better test coverage, and faster time-to-market. AI-driven tools make automated testing smarter, more reliable, and more adaptable to the ever-changing software landscape.

As you can see AI in software testing takes many forms: generative AI for test scripts, natural language processing for, vision and even audio processing, machine learning, data science, etc. These are all mixed. The good news is that testing using artificial intelligence doesn't require you to have deep understanding of algorythms, and tech, and types of ML learning. You just need to choose the right AI testing tools... and not fall for the lies.

AI Tools: Optimize Testing but Don’t Believe Everything They Promise

We’ve been in the market for AI tools for over a year, searching for a partner that truly enhances our automated testing on both front and back ends. Many tools we encountered used AI as a buzzword without offering real value. It was frustrating to see flashy promises without substance.

Then we found Virtuoso AI. It stood out from the rest.

“With Virtuoso, our trained professionals create test suites effortlessly. These are structured logically, maintaining reusability and being user-centric. Once we establish a baseline, maintaining test suites becomes straightforward, even as new releases come in. Regression suites run quickly and efficiently.”

Bruce Mason, UK and Delivery Director

Key areas of Virtuoso’s AI product include

Codeless automation. We can set up tests just by describing what they need to do. No coding necessary, which means quicker setup and easier changes.

Functional UI and end-to-end testing. It covers everything from button clicks to complete user flows. This ensures your app works well in real-world scenarios, not just in theory.

AI and ML integration. AI learns from your tests. It gets smarter over time, improving test accuracy and reducing manual adjustments.

Cross-browser testing and API integration. With this tool we can test how your app works across different web browsers and integrates API checks. This means thorough testing in diverse environments – a must for consistent user experience.

Other AI Tools for Testing

Besides Virtuoso AI, here are a few other notable artificial intelligence software testing tools available on the market:

• Applitools. Specializes in visual AI testing, offering tools for automated visual validation and visual UI testing.
• Testim. Uses machine learning to speed up the creation, execution, and maintenance of automated tests.
• Mabl. Provides an AI-driven testing platform that integrates with CI/CD pipelines, focusing on end-to-end testing.
• Functionize: Combines natural language processing and machine learning to create and maintain test cases with minimal human intervention.
• Sealights: Focuses on quality analytics and continuous testing, offering insights into test coverage and potential risk areas.

When evaluating these tools and testing activities they cover, remember to check their true AI capabilities, scalability, integration, and support systems to ensure they meet your needs.

But let’s not ignore the broader market. There are many AI tools available, each with its own strengths and weaknesses. Here’s what to consider when evaluating them:

• True AI capabilities. Look beyond the buzzwords. Ensure the tool offers genuine AI-driven features, not just automated scripts rebranded as AI.
• Scalability. Can the tool handle large-scale projects? It should adapt to your growing needs without performance issues.
• Integration. Check how well the tool integrates with your existing systems and workflows. Seamless integration is crucial for efficiency.
• Support and Community. A strong support system and an active user community can make a significant difference. Look for tools with responsive support teams and extensive documentation.

Choosing the right AI tool for testing is critical. It’s easy to get caught up in marketing hype. Stay focused on what truly matters: real, impactful features that improve your testing process. Our experience with Virtuoso has been positive, but it’s essential to do your own research and find the best fit for your needs.

In summary, while AI tools can optimize testing, be cautious and discerning. Not all tools deliver on their promises. Seek out those that offer genuine innovation and practical benefits.

What are the Disadvantages of AI in Software Testing?
If you feel like the previous part confirms that you may be out of work… soon, don’t sell yourself short, at least for now. Here are the limitations AI has and will have for a considerable amount of time.

1) Lacks creativity. AI for software testing algorithms experience big problems generating test cases that consider edge cases or unexpected scenarios. They need help with inconsistencies and corner situations.
2) Depends on training data. Don’t forget — artificial intelligence is nothing else but an algorithm, a mathematical model being fed data to operate. It is not a force of nature or a subject for natural development. Thus, the quality of test cases generated by AI depends on the quality of the data used to train the algorithms, which can be limited or biased.
3) Needs “perfect conditions.” I bet you’ve been there — the project documentation is next to none, use cases are vague and unrealistic, and you just squeeze information out of your client. AI can’t do that. The quality of its work will be exactly as good or bad as the quality of the input and context turned into quantifiable data. Do you receive lots of that at the beginning of your QA projects?
4) Has limited understanding of the software. We tend to bestow superpowers on AI and its understanding of the world. In fact, it is truly very limited for now. May not have a deep understanding of the software being tested, which could result in missing important scenarios or defects.
5) Requires skilled professionals to operate. For example, integrating a testing strategy with AI-powered CI/CD pipelines can be complex to set up, maintain, and troubleshoot, as it requires advanced technical skills and knowledge. Tried and true methods we use now may, for years, stay much cheaper and easier to maintain.

How AI-Based Software Testing Threatens Users and Your Business

There is a difference between what AI can’t do well and what can go wrong even if it does its job perfectly. Let’s dig into the threats related to testing artificial intelligence can take over.

• Bias in prioritization and lack of transparency. It is increasingly difficult to comprehend how algorithms are making prioritization decisions, which makes it difficult to ensure that tests are being prioritized in an ethical and fair manner. Biases can influence artificial intelligence models/tools in the data used to train them, which could result in skewed test prioritization.

Example. Suppose the training data contains a bias, such as a disproportionate number of test cases from a particular demographic group. In that case, the algorithm may prioritize tests in a way that unfairly favors or disadvantages certain groups. For example, the training data contains more test cases from men than women. The AI tool may assume that men are the primary users of the software and women are secondary users. This could result in unfair or discriminatory prioritization of tests, which could negatively impact the quality of the software for underrepresented groups.

• Overreliance on artificial intelligence in software testing. Lack of human decision-making reduces creativity in testing approaches, pushes edge cases aside, and, in the end, may cause more harm than good. Lack of human oversight can result in incorrect test results and missed bugs. Increased human oversight may lead to maintenance overheads.

Example. If the team relies solely on AI-powered test automation tools, they may miss important defects that could have significant impacts on the software’s functionality and user experience. The human eye catches inconsistencies using the entire background of using similar solutions. Artificial intelligence only relies on limited data and mathematical models. The more advanced this tech gets, the more difficult it is to check the results’ validity, and the riskier is overreliance. This overreliance can lead to a false sense of security and result in software releases with unanticipated defects and issues.

• Data security-related risks. Test data often contains sensitive personal, confidential, and proprietary information. Using AI for test data management may increase the risk of data breaches or privacy violations.

Example. Amazon changed the rules it’s coders and testers should follow when using AI-generated prompts because of the alleged data security breach. It is stipulated that ChatGPT has responded in a way suggesting it had access to internal Amazon data and shared it with users worldwide upon request.

So, What Will Happen to AI in Testing?

What is the future of software testing with AI?

We don’t know.

You don’t know.

Our partners at Virtuoso AI don’t know.

We can guess the general direction —

• Manual testers will get more into prompting and generate test scripts that will allow more coverage with fewer motions; Expert manual testers will also be more valued for human touch and human eye checking after AI testing tools;
• Test automation frameworks will be almost 100% driven by AI;
• Continuous testing will become more affordable than ever;
• “We needa large number of test cases” trend will be overrun by priritization in testing and monitoring;
• soon there will be tools for almost any testing needs, but only the most efficient and affordable solutions will survive the competition. AI is transforming how we do software development and testing.

AI is transforming how we do software development and testing.

If you are a manual QA beginner — you better hurry and invest in your skills. The less expert and the easier to automate tasks you do now, the faster algorithms will come after your job. In the end, here is what Chat GPT thinks of it:

In our company, we started to apply AI-based tools for test automation back in 2022 and continue adopting new tech with new partners — Virtuoso, Google, Amazon, etc.

Will it be enough to stay relevant and efficient?

We definitely hope so. AI can help, but the software testing process is much more complex than just applying new tricks.

This content originally appeared on DEV Community and was authored by TestFort