LaunchDarkly turned skepticism into a new engineering standard by using Meticulous to reduce manual engineering work and better leverage AI coding tools.
“Meticulous helps us simultaneously move fast and raise the quality bar. If we didn't have Meticulous, it would be so much more difficult to get to that place.”
—Zach Davis, Principal Engineer, LaunchDarkly
80%
reduction in major frontend incidents
THE Challenge
25% of Fortune 500 companies use LaunchDarkly to measure impact, optimize features and release with confidence. But with a growing engineering team rapidly adopting new AI code generation tools, maintaining their own high-quality releases became a challenge.
Profile
Industry
Developer Tooling
Age of codebase
12 years
Engineers
300+
“We’re moving into an era where code generation is relatively inexpensive, but that means code review and the feedback loop become the bottleneck.”
As AI made generating code easier, the real bottleneck in software development shifted to validation and code review. LaunchDarkly had end-to-end integration tests, but like most teams, they couldn’t cover every edge case - and expanding coverage always came with the tradeoff of spending more time maintaining tests.
Hiring more engineers or dedicating more time to QA was expensive and inefficient, and manual tests couldn’t scale. For LaunchDarkly, automated testing became the only way to ensure coverage without slowing down development.
“I’m trying to figure out how we simultaneously move faster, increase confidence, and reduce the number of bugs we ship to production,” Zach explained.
THE Solution
After hearing about Meticulous from a friend, the team began implementing the platform. On the very first day of the rollout, they had a production incident.
When Zach checked Meticulous, he realised it had been the only tool to flag the issue.
“We realized it would've prevented us from having a production incident if we had it in place an hour earlier. That’s when my initial confidence shifted,” he said.
Since adopting Meticulous, LaunchDarkly has seen an 80% reduction in frontend incidents.
The team’s new goal is simple: zero frontend regressions.
“Meticulous has enabled us to ship way fewer regressions to customers,” Zach said. “My expectation now is that our baseline should be zero production incidents caused by frontend changes.”
The team’s increase in confidence comes not just from seeing Meticulous catch bugs, but from seeing it catch issues that would have slipped through PR reviews or existing tests and tools. Knowing that every edge case is being reviewed gives engineers peace of mind.
“The number of bugs Meticulous catches builds trust across the team and increases confidence in every change,” Zach said.
Zach calls it a fallacy of engineering management to think an organization will necessarily achieve more or move faster with a higher headcount. Instead, engineering leaders should focus on efficiency. By investing in a PR feedback cycle with Meticulous, LaunchDarkly achieves greater efficiency, regardless of who wrote a particular change.
Engineers now spend less time on PR reviews and manual testing - and more time shipping. The platform allows them to safely integrate AI-generated code without worrying about unintended regressions.
Monthly Test Run Stats
52,310
user flows mapped
3,754
test runs
15,495,718
screenshots
“I think the reason Meticulous helps us move faster is it increases confidence when we're making changes or, in the crazy new world that we live in, where AI agents are making changes.”
The impact
For example, LaunchDarkly maintains a shared design system in a separate repository that is regularly pulled into the core application as a dependency. Historically, small changes to the design system would cause unpredictable effects across the product, breaking overrides or obscure edge cases that were impossible to validate manually. With Meticulous, any updates to the design system are automatically validated across the entire application.
Meticulous has also empowered engineers to confidently tackle large-scale refactor work. Recently, the team used it to complete several major frontend refactors to address slow tooling and dependency sprawl. Under the previous testing framework, this work would often carry a high risk of regressions.
Zach believes that without fast, reliable validation, AI code generation slows teams down. If every change requires heavy manual review or QA sign-off, the velocity gains disappear.
“So if you're paying gobs of money for code generation, but your PR review is slow, or you have to have a QA team come behind and check every change, then you're not really speeding things up that much.” said Zach.
Meticulous closes that gap by accelerating the PR feedback loop, regardless of whether code is written by humans or AI.
For Zach, Meticulous is essential to how LaunchDarkly builds software in the AI era - enabling teams to ship faster, confidently adopt codegen tools, and maintain high quality across hundreds of real product flows.
