How GenAI Is Transforming the Last Frontier of Development

Mention “testing,” and you can almost hear a virtual groan among code warriors.

It’s not that they do not see it as essential; the truth is that software testing takes a lot of effort. And for a developer who is primarily motivated to create something new, they have no desire to maintain an old code base, write test scripts, and fix bugs.

But the world of software development has changed dramatically. “25 years ago, testing meant writing scripts. You needed specialized coders just to create test automation,” commented Damien Wong, senior vice president for APAC at U.S.-based Tricentis. “That was less of an issue then because there were fewer applications in the landscape, less integrations, and the changes happened less frequently.”

Fast forward to today, we are seeing an explosion of applications with significant “ripple effects” for ill-defined or untested code. With so many software applications with complex dependencies, some even touching infrastructure through Infrastructure as Code (IaC), and the real threat of supply chain attacks hanging over the heads of many enterprises' software development teams, the repercussions have mushroomed.

GenAI was supposed to help. “Today, with the help of GenAI tooling and large language models, you will see that the speed at which code is being developed has accelerated exponentially,” Wong explains.

Until it didn’t. GenAI augmented past challenges with new ones, such as hallucinations and inefficient code. “Where we are seeing bottlenecks is in the validation of that code… code that is auto-generated by some of these GenAI tools introduce errors and all vulnerabilities,” Wong adds.

This has made software testing a major bottleneck to software modernization and transformation. It is also where a company like Tricentis — whose customers include ANZ, Dolby, Heineken, Hong Kong Jockey Club, IHG, Jollibee, Singapore Pools, and Woolsworth — is looking to plug the gap with its AI-powered testing.

And in today’s hyper-accelerated software development world, such tools, which include the Tricentis Tosca and Testim, are becoming part of a critical survival strategy for modern DevOps teams.

Beyond script monkeys: The AI testing revolution

Tricentis’s approach uses GenAI-augmented, model-based, and codeless testing that does not begin with the code itself. “So, model-based test automation effectively abstracts away the business processes of the business model from the underlying technology,” Wong says. “If there are changes to the application, [these are] automatically propagated to the hundreds of thousands of test cases that are automated.”

Such an approach is gaining traction among developers who have traditionally seen testing as hindering their creativity and wasting precious time. A recent study commissioned by Tricentis with Techstrong Research revealed a stunning insight: 60% of DevOps professionals see the greatest AI value in testing, not coding.

But that is not the most revolutionary aspect. GenAI is also breaking down technical barriers. Non-technical team members can now participate directly in testing processes. This democratization of the testing function means those with domain expertise can now get directly involved in the software testing phase without having to learn scripting as they will know best whether the software outcomes or model behavior is on track.

“In the past, if a business user wanted test scripts, they’d speak with the teams that are responsible for engineering test automation,” Wong explains. “Now, we remove that barrier entirely.”

Another artifact of this approach is transforming modern software testing into one that’s proactive.

For example, Tricentis' vision is to create test frameworks from mere sketches. Wong recalls a mind-blowing example: “We had teams draw application prototypes on flip charts, photograph and scan them, and immediately generate test frameworks — before a single line of code was written.”

Tricentis goes further with using metadata instead of raw data to ensure that each customer's intellectual property and privacy remains sacred. “Now, metadata is very different from data; it's kind of what we always differentiate on. With metadata, you can't reverse engineer your software.”

The company is also shoring up its product line for mobile B2B enterprise apps as they become critical in operations, not just in offices. It recently acquired Waldo, a SaaS-based, no-code, zero-footprint mobile test automation platform, to venture into mobile test authoring and execution.

Changing the script with GenAI

Wong sees tools like Tricentis Tosca Copilot, Testim Copilot, and more recently, the qTest Copilot changing the testing game forever. More importantly, it makes the entire process more scalable. “They allow QA and developer teams to greatly accelerate software delivery,” he adds. “Think about the intelligent heart of an autonomous car. So we have the same thing [with GenAI] but for test management.”

Last year’s SeaLights acquisition takes Tricentis’s value proposition further. It allows the company to provide AI-enabled quality intelligence “beyond SAP environments and into both custom and packaged applications, including test impact analysis, quality risk management, root cause analysis, and support across all programming languages,” said the company statement. 

Essentially, software development teams can now understand where new code changes “can have an impact, assess the risk of new code changes, determine if there are code coverage gaps, and resolve bugs introduced with new code changes all in a continuous automation testing cycle.”

This is important for larger enterprises that are often saddled with legacy codebases.  “We often talk about GenAI and digital natives, but they have very little legacy technical debt. Whereas you talk to a bank or an airline, they have a lot of legacy. And even though they are building web front ends, mobile front ends, etc., they still have to deal with mainframes and legacy client-server systems,” says Wong.

The acquisition, for example, allows enterprise quality assurance teams to view the changes they need to make with a new regulatory requirement. That’s agility and resilience at the code level.

But while GenAI may make software testing more seamless and accessible, it is prone to hallucinations. While you can tolerate these in a written report, hallucinations can render software testing biased and faulty. It may result in software that may appear fully tested, but it is really full of holes.

It’s why Wong is adamant about avoiding the pitfalls of generic large language models. “We don't use vanilla LLMs,” he emphasizes. “We contextualize AI specifically for testing environments, ensuring privacy, reducing bias, and minimizing hallucinations.”

He is also clear that Tricentis’ focus on GenAi isn't about replacing human expertise — it's augmenting it. “Our tools are co-pilots, not autonomous systems,” Wong says. “We expect human review, but we're reducing manual effort by 80-90%.”

It’s about survival

As digital transformation accelerates and GenAI continues to reshape development practices, one thing becomes clear: the future of software quality isn't about catching bugs anymore. It's really about preventing them before they're even born.

That’s important in today’s increasingly complex technological landscapes — with some enterprises running over 1,000 interconnected applications. This makes GenAI-powered testing more about survival, as a single oversight or bad line of code can delete years of reputation.

Game on, code warriors!

Image credit: iStockphoto/Suratsak Noikerdmee