Tricentis looks to agentic AI as firms risk losses from untested code

Organisations across the Asia-Pacific region are increasingly sacrificing software quality for speed, a trend that Tricentis senior vice-president for the region Damien Wong described as “scary”.

In an interview with Computer Weekly in Singapore, Wong cited the company’s recent research on transforming software quality, which found that nearly half of the organisations surveyed admitted to releasing code without testing it.

“The pressure to deliver fast is sometimes driving this behaviour to cut corners and roll out a piece of software before it is thoroughly tested,” Wong said. “Speed is being pushed very hard by businesses.”

The financial implications of untested codes can be significant. According to Wong, organisations could face average losses of between $500,000 and $5m due to software defects and outages.

To address this, Tricentis has been investing heavily in agentic test automation. Unlike earlier iterations of artificial intelligence (AI)-powered software testing assistants that function more as co-pilots, AI agents have a level of autonomy where they are capable of generating test assets, executing them, and analysing results with minimal human intervention.

“Imagine you have a virtual performance engineer now, and you can use natural language to instruct it,” Wong said. “You can say, ‘Test the system with a maximum of 1,000 concurrent users, return the results, and make some recommendations to improve its performance'.”

Wong noted that the use of automated software testing tools does not mean human testers will lose their jobs. Instead, they will be more productive and be valued much more than before. “Imagine one AI-enabled test engineer who is highly competent. They’re going to be able to do the job of 10 or 20 engineers to cope with the speed of code development that’s going through the roof.”

The demand for automated software testing tools is reflected in the company’s regional performance. Wong revealed that Tricentis’s Asia-Pacific business grew by 42% year-on-year in financial year 2024, outpacing the company’s global growth rate of 28%.

The company’s growth has been fuelled by two key trends in the region: application modernisation and generative AI-driven digital innovation. As legacy “monolithic and brittle” applications are updated to cloud-native stacks, the complexity of testing increases exponentially, Wong said.

“You have hundreds, thousands, or even tens of thousands of applications running all over your organisation,” he said. “You make one change in one system or application, and there are potential impacts to other systems and applications upstream or downstream.”

Wong noted that agentic testing – and the broader realm of software quality assurance – is also about visibility. Pointing to the company’s acquisition of SeaLights, an AI-powered software quality intelligence platform, he explained how AI can prevent disasters similar to the 2024 CrowdStrike outage, which was caused by a content configuration update that was not fully tested.

Wong said the SeaLights platform can constantly monitor application metadata for code changes and correlate them with testing activity. “For example, if you’ve changed an API [application programming interface] and checked in the code, what sorts of tests have you run? If no tests have been done, it intelligently tells you that this is a problem. You’ll never be caught flat-footed if something goes wrong.”

As organisations rush to integrate generative AI into their products, they are also grappling with how to validate non-deterministic large language models that produce varied outputs from the same input, making traditional validation methods inadequate.

“There is no right or wrong outcome, but there could be things that are completely unacceptable. So, you want to make sure that you’re able to define the guardrails and ensure you stay within them,” Wong said.

While the market is still maturing in this area, Wong said a recent Tricentis masterclass in Singapore on testing intelligent systems was oversubscribed, indicating that organisations are anxious about AI risks and compliance. However, Wong said that for now, most customers are using AI to test traditional, deterministic systems.

Tricentis’s tools can be consumed through the public cloud and on-premises. While the company has established cloud regions in Australia and Japan, with plans to do so in Singapore, many regulated industries still prefer on-premises or hybrid solutions, Wong noted.

“In Asia-Pacific, the trends have been mixed,” he said. “There was a huge trend towards going full public cloud, but there has been a bit of pullback. Many of the heavily regulated customers we work with elect to deploy our software on-premises out of security concerns,” he added.

That said, Wong assured customers who use Tricentis’s cloud platform that the company does not have access to their source code. “We can see the metadata associated with your software builds to determine what modules have been changed. But we don’t exactly know the code in them, so that insulates people from the fear that their code might be taken or stolen from them,” he said.