Software testing startup Arato gets $10M to stop businesses deploying AI systems blind

Israeli startup Arato Software Ltd. is developing tools for developers to test and evaluate their artificial intelligence applications, and it has just gotten $10 million in seed funding to pursue that goal.

Today’s round was led by TLV Partners and saw participation from Jibe Ventures. Former VMware Chief Executive and now Andreessen Horowitz partner Raghu Raghuram and former Intuit Inc. Chief Technology Officer Marianna Tessel also participated in the round.

Arato, founded in 2024, has built a platform that aims to simulate how users might interact with AI applications, so that organizations can try to spot potential failures before new systems reach production deployment. Using Arato’s tools, developers can simulate thousands of scenarios using text, image, voice and business data before analyzing the results to spot recurring issues with their apps, identify risks and highlight areas for improvement.

This kind of platform is urgently needed, because AI systems are quite unlike traditional applications. They have the potential to generate a much wider range of unpredictable responses based on what the user inputs, making them more difficult to test before they can be safely released. With Arato’s platforms, organizations get the reassurance of an additional validation layer that helps them understand how new AI applications are likely to perform in the wild and ensure they adhere to the security and regulatory requirements that govern them.

Arato’s evaluation tools are designed to be used both before deployment and also continuously once an AI application reaches production. In this way, it provides ongoing visibility into the performance of AI systems, how they’re affecting user experiences, and where improvements may be required.

Co-founder and Chief Executive Shahar Erez said the reality of AI today is that the technology hasn’t yet been perfected, but organizations are deploying it anyway because they believe they can’t afford not to do so. “The question is how often will it fail, in which situations, and what the potential damage could be to users and to the business,” he explained.

For many developers, the challenge is to ensure that AI applications solve business problems and create genuine value for their users, while minimizing the downsides that come with an untested technology. “Arato makes it possible to measure that systematically before customers discover problems themselves,” Erez said.

According to Erez, Arato already has dozens of customers using its tools, including enterprise software providers, e-commerce companies, financial institutions, insurance firms and industrial businesses. One of those customers is a global industrial firm that deployed an AI assistant for field technicians.

Using Arato’s software, it reduced the validation times for new updates to that AI assistant from three months to a matter of days. It reduced the manual effort that goes into this by around 80%, and its continued use is expected to save that company around $5 million in costs over the next three years.

Erez founded Arato alongside CTO Hilik Paz, who previously helped him to found the Israeli workforce management software startup Stoke Talent Ltd., which was acquired by Fiverr International Ltd. for $95 million back in 2021. Arato’s third co-founder is Vice President of Research and Development Tal Salmona, who previously worked with Erez and Paz at Mercury Interactive Inc. and VMware Inc.

TLV Partners Managing Partner Eitan Bek said most businesses are blindly deploying AI systems into business-critical workflows with little idea of how they’re going to behave when they’re up and running in production. “Arato is addressing a problem that will become increasingly central as AI adoption scales,” he said. “It’s giving product, engineering and business teams the evidence they need to know whether an AI application is ready for production.”

Photo: Eyal Toueg