Applause survey finds AI use rises but defects grow

Applause has found that 92% of organisations now use AI in software testing. The survey also found that 29% of teams reported a rise in the number or severity of defects.

The company's 2026 functional testing study points to a sharp increase in AI use in quality assurance and software development, with only 8% of respondents saying they do not use AI in any part of testing. Yet the figures suggest that wider adoption has not reduced testing problems for many teams.

Another 15% of respondents said both the number and severity of functional testing issues had increased. At the same time, 89% said AI had changed how they test applications and digital experiences.

Human role

The findings suggest that many organisations still see a clear role for people in testing. The survey found that 86% of respondents said human involvement is extremely important to functional testing.

More specifically, 57% said humans remain critical for qualitative feedback through peer review. The same share said people are equally important in designing test strategies based on real-world user behaviour.

AI use appears to be more established in software development than in testing, although the gap is narrowing. Overall, 79% of respondents said AI had a significant or moderate impact on development, compared with 72% who said the same for testing.

Only 3% said AI had no impact at all on development, while 9% said it had no impact on quality assurance. The results were based on a global survey of members of the uTest community and other software development, QA, product, AI and data science professionals, along with interviews with technology leaders.

Common uses

Among development teams, 62% said they use AI coding assistants to generate or complete code in editors. A further 59% said they use AI for code review and documentation.

In testing, the most common uses were creating test cases, cited by 65% of respondents, writing test automation scripts at 62%, and identifying or addressing coverage gaps at 48%.

The study also suggests companies are starting to formalise internal rules for these tools. Some 66% of organisations said they now have documented policies covering AI use in development and testing, although only 20% described those guidelines as clear and robust.

Tacita Morway, Chief Technology Officer at Applause, said the speed of software delivery had changed the demands on testing teams. "AI has accelerated the development process, and that's changing what teams need to test and how they need to test it," Morway said. "Traditional automated testing answers the question: can this task be completed? A human tester answers a harder one: could a real person work out how to do this, and get it done? Teams are giving up that second question as release cycles get faster than people can keep up with, and I think that accounts for a lot of the defect increase we're seeing. The work now is to get that kind of human testing coverage back, but at machine speed. That takes AI, automation and people together."

The report also examined the implications for newer AI systems that act with greater autonomy. Morway said these systems require different forms of assessment because the same request may not produce the same chain of actions each time.

"This gets harder with evaluating agentic systems that make decisions and take action on their own. With an agent, there's no fixed set of steps to check - the same request can take a different path every time. The failures that matter most happen when the system is effectively making a judgment call and gets it wrong - inaccurate financial transactions, exposed sensitive information, dangerous guidance. Those carry real consequences for organisations and their customers, and you can't catch them by confirming a script still passes. Systems this powerful and this unpredictable need a materially different approach to evaluation," Morway said.

The data highlights a tension for software teams adopting AI tools across the development cycle. While automation is becoming more common in writing code, reviewing it and preparing tests, defect levels remain a concern for a sizeable minority of teams.

Morway said organisations need to preserve user-centred testing even as development speeds increase. "Shipping faster only helps if you can still trust what you're shipping. That means testing that's grounded in real-world context and backed by human expertise," Morway said. "Our survey showed 86% of respondents understand that human involvement is essential to functional testing, but many of those organisations are struggling to preserve that human perspective while supporting faster development. Traditional automation doesn't solve that. Intelligent automation does - grounded in the product, the user and the risks that matter. Human expertise helps shape what gets tested and how, and fills the gaps where automation can't or shouldn't."