Leapwork removes code verification from AI agents in favor of automated software development

Danish artificial intelligence startout Leapwork ApS has announced the launch of the Automatic Continuous Validation Platform, which aims to help businesses keep pace with AI software development.

The rise of the AI ​​”vibe coding” means that software development has accelerated to new heights, allowing businesses to roll out new code faster than ever before. That’s great for innovation, but it greatly increases the risk of teams accidentally breaking something.

Leapwork’s new Continuous Validation Platform is designed to prevent that. Keeping up with the volume of AI-generated code being shipped, it’s also an automated “agent” platform that aims to provide a sort of safety net for engineering teams. The idea is to ensure that every piece of code, whether written by humans or AI agents, is validated at every step of the development life cycle and everywhere.

The startup aims to widen the “trust gap” in software engineering. Citing its own research, it says 88% of software developers now prioritize AI. That has a big and negative impact on their quality assurance budget, which has ballooned from 35% to 40% of their total spend, on average.

The problem is that traditional code inspection tools can’t keep up with the compressed release cycles of AI-generated software. Bots can generate new code in minutes, compared to hours for human developers, yet their results can be devastating. So when AI builds software, more emphasis needs to be placed on validation – proving that the code works in a prescribed and testable way so it can be used safely in production.

CEO Kenneth Ziegler said the Continuous Validation Platform is “application-agnostic,” meaning it can be used to validate code across modern web, mobile and legacy systems like Oracle and SAP. The platform consists of three main components, including Leapwork Studio, which is a code-free engine with “AI Automation Blocks” that aims to ensure that AI results are correct and visualize data in the same way as humans.

The second part is Leapfrog Performance, which includes performance and testing of the application interface in the software development workflow. With this, engineering teams can import their existing JMeter scripts into their workflows and use them to model real-world traffic scenarios.

Ultimately, Leapwork AI Studio is intended to be the “agent” heart of the platform, built to validate code written with tools like Claude Code and Cursor. It is based on an intelligent knowledge base that imports the newly generated code and designs the test programs and documentation required for verification.

Ziegler said Leapwork wants to be the “defining foundation” for the era of automated software development. He said engineering teams are excited about agent automation, but also face “unprecedented concerns” about the stability of their AI-generated code.

“Continuous validation is expensive and tricky to obtain without background expertise, history and experience,” he said. “With the Leapwork platform, our customers have access to the speed and intelligence of AI-native automation and the proven reliability that businesses need to meet their quality and governance needs.”

Photo: SiliconANGLE/Meta AI