Against the backdrop of high-profile outages disrupting global infrastructure, only 16% of businesses think their current testing practices are efficient.

Research from Leapwork shows that the surge in AI adoption may be exacerbating this issue, as they surveyed 401 senior/technical professionals in the US and the UK.

Although 85% of total respondents have integrated AI apps into tech stacks in the past year, over two thirds (68%) have experienced issues with their performance, accuracy, and reliability.

The lack of sufficient testing processes is having a detrimental impact on the usability of AI apps, which are becoming more and more widespread.

For C-suites, which accounted for half of the overall survey, this perception is even higher – nearly three quarters (73%) of executives have experienced issues with AI apps, compared to 62% for software engineering or technical leads, which made up the other half of the sample.

The need for software testing should therefore be even more apparent, especially considering the global catastrophe that follows major outages, as seen in the Crowdstrike outage this August.

“For all its advancements, AI has limitations, and I think people are coming around to that fact pretty quickly,” says Robert Salesas, CTO at Leapwork.

“The rapid automation enabled by AI can dramatically increase output, but without thorough testing, this could also lead to more software vulnerabilities, especially in untested applications. It makes sense that C-Suite executives would be especially sensitive to this because of the implications for customer experience and negative publicity.

“There’s an opportunity here for cross-industry collaboration to ensure more testing tools are up to scratch for the challenges of the modern world where AI apps are more and more widespread.”

The rapid advancement of AI apps has presented challenges when it comes to onboarding them into enterprises’ workflow.

Specifically, the most common bugs in AI apps identified by respondents were integration failures (21%), and security vulnerabilities (23%).

These were followed by resistance to change in the organisation (20%), inconsistent performance and reliability of AI applications (19%), and managing the rapid pace of AI advancements and updates (19%).

Recommended reading

• CrowdStrike Incident | Lessons Learned and Key Questions Answered
• CrowdStrike Reveals Global Outage Details
• Top 5 Strategic Tech Trends in Software Engineering for 2024

Consequently, most organisations (77%) now see testing of AI as essential, but notable gaps exist in testing resources and practices.

Nearly a quarter of organisations (24%) do not hae a dedicated team or individual responsible for testing AI apps and over a quarter (26%) do not have a commercial testing platform. Further, nearly a third (30%) believe their current testing processes cannot ensure reliable AI apps.

“There have been too many outages this year alone, many of which affected millions of customers for big brands. We’ve been given a wake-up call no one can ignore,” says Christian Brink Frederiksen, CEO of Leapwork.

“What makes digital infrastructure today so tricky to test is the copious amount of complex, interconnected applications. A tiny error in one application could have a monumental cascading effect and shut down businesses. Whether big or small, all updates need appropriate testing, but many businesses have outdated, siloed approaches. It shouldn’t be about testing one individual app – it should be about testing the entire user journey.”