CloudBees Smart Tests Goes GA: Taming the Flood of AI-Generated Pull Requests in CI

A New Bottleneck: Validating Code, Not Writing It

For years, the software delivery bottleneck was writing code. In 2026, it's validating it. CloudBees just made that shift official with the general availability of CloudBees Smart Tests, its AI-driven test intelligence platform for CI/CD, now open to all customers. The timing is deliberate: CloudBees points to a landscape where roughly 41% of all code is now AI-generated and more than 80% of developers use AI coding tools daily, a surge that is expanding regression suites and slowing feedback loops across enterprise pipelines.

What Smart Tests Actually Does

Under the hood, Smart Tests applies machine-learning-based Predictive Test Selection alongside failure pattern analysis to figure out which tests actually matter for a given code change, rather than re-running an entire suite every time a pull request lands. That's the same core capability Launchable, the AI test-intelligence startup, built before it was folded into CloudBees. Instead of brute-forcing full regression runs, the platform triages test failures, flags likely-flaky tests, and automates root-cause analysis so engineers spend less time babysitting CI and more time shipping.

CloudBees is backing this up with numbers from early enterprise deployments: teams report up to 80% faster test execution, 40% shorter build times, and roughly 2,000 developer hours saved per month at scale. In one customer example, GoCardless used Smart Tests to intelligently subset its suite, focusing runs on the tests most likely to fail and executing those first, cutting both wall-clock test time and cloud compute spend in the process.

The Launchable Backstory

Smart Tests didn't appear out of nowhere. CloudBees acquired Launchable, the predictive test selection startup founded in 2019, on August 7, 2024, in a deal that also marked the return of Jenkins creator Kohsuke Kawaguchi and co-founder Harpreet Singh to CloudBees. At the time, CloudBees framed the acquisition as bringing "the first AI-augmented Test Intelligence capability to any DevSecOps platform" - a bet that AI-assisted coding would need an equally AI-assisted testing layer to keep pace. Eighteen months on, that bet looks prescient: the volume of AI-generated pull requests has made naive full-suite regression testing untenable for teams shipping multiple times a day.

Why This Matters for QA Teams

For test engineers, the GA launch is another signal that predictive test selection is moving from a niche optimization to table stakes. As AI code generation tools push commit volume higher, CI queues that used to take minutes can balloon into hours unless something intelligently prunes what gets re-tested. Teams evaluating CI/CD tooling in 2026 should expect "which tests actually need to run" to become as standard a pipeline question as "which tests failed." Whether via CloudBees Smart Tests or rival test-impact-analysis tools, the underlying pressure is the same: more code is being written faster than ever, and testing strategy has to get smarter, not just bigger, to keep up.

What to Watch Next

Expect competitors in the test-impact-analysis and observability-driven testing space to respond with similar GA announcements and case-study data of their own in the coming months, as CI pipeline cost and speed become a board-level conversation alongside AI coding adoption metrics.