CloudBees buys co-founder's AI test automation tools startup

QA and test optimization: GenAI's next big moment?

Launchable lists customers on its website including UKG, BMW, Sony, Vitess, Delphix and Infosys; Kawaguchi declined to disclose the total number of customers. The startup also has a decidedly CloudBees pedigree even beyond its co-founders: three CloudBees executives, including CloudBees co-founder, board member and chief strategy officer Sacha Labourey, are listed among its advisors and investors.

Two Atlassian executives are also on that list -- Singh worked there for a year as general manager for Bitbucket Cloud before co-founding Launchable. But industry analysts say Launchable seemed destined for a CloudBees reunion, given it's already built on Jenkins.

With DevOps code velocity increasing with AI-based coding, testing, QA and validation has all but taken a back seat.

"It was only a matter of time before they merged back together to create more value for DevOps teams," said Andy Thurai, an analyst at Constellation Research. "With DevOps code velocity increasing with AI-based coding, testing, QA and validation has all but taken a back seat."

This has led to untested code changes rolled into production too hastily, with the CrowdStrike outage being one high-profile example, Thurai said.

AI code generation also encompasses automatically generated test code, but Launchable goes a step further in trying to make all that test code more manageable, said Jason English, an analyst at Intellyx.

"There's a whole lot of aspects of even just being a QA engineer now where there's just too much to get your arms around," he said. "If you're not automating 80% of your test code, you're probably missing 80% of your bugs. AI is really good at that sort of systematic thinking and speaking the language of development and testing."

Launchable isn't alone in spying the opportunity to bring AI optimization to test automation tools -- its competitors include Tricentis, through its recent SeaLights acquisition and other startups such as LambdaTest, Diffblue, Testaify and SmartBear.

CloudBees plans self-tuning expansion

With Jenkins integration for Launchable already available, CloudBees will turn its focus to expanding Launchable's automation features to the rest of its DevSecOps platform, according to Shawn Ahmed, chief product officer at CloudBees.

Shawn Ahmed

"The idea around triaging tests, I want to be able to use it for not just QA but for build [and] progressive deployment," Ahmed said. "What I'm really trying to do is make our DevSecOps platform more semi-autonomous."

This idea is also being pursued by CloudBees competitors, most notably GitLab, which offers root cause analysis on CI/CD pipeline failures through its GitLab Duo AI product. Atlassian partners market AI test automation add-ons through its marketplace, and its Atlassian Intelligence for Bitbucket cloud can automatically create pull requests, including generating, summarizing or transforming their descriptions and comments. Observability vendors such as Datadog and Dynatrace have added AI-driven features for DevOps pipeline automation and source code analysis, and Datadog has reportedly been in acquisition talks with GitLab.

CloudBees and Launchable might face questions about AI governance, as Launchable partners with OpenAI on its back end. Launchable has also earned SOC II certification and allows users to control how their data is used, Kawaguchi said.

"We prioritize the security and confidentiality of our customer data through robust encryption and security audits," he said. "We maintain transparency in our data practices, giving users full control over their information with clear privacy policies and consent mechanisms."

Still, the risks of generative AI are weighing heavily on enterprise IT organizations, according to Constellation's Thurai.

"We are reaching a stage in which AI can help code, create test plans, help test and move code to the production environment. [This] should all be taken seriously with proper manual oversight, governance and security every step of the way," he said. "Continuing [to add AI to] the full software supply chain can lead to the unknown unknowns, issues that were never encountered before [and] given limited human involvement. It might become even more complicated to do root cause analysis, thus defeating the purpose of this whole automation and efficiency process."

Beth Pariseau, senior news writer for TechTarget Editorial, is an award-winning veteran of IT journalism covering DevOps. Have a tip? Email her or reach out @PariseauTT.