Snowflake, the cloud-based data platform, is known for its fast-paced growth and strong engineering culture. With hundreds of developers working on highly-complex projects, the company faced an ongoing challenge: how to better streamline code review workflows, reduce bottlenecks, and empower engineers to ship high-quality code quickly.
Graphite entered Snowflake through a grassroots adopter: one senior engineer who had used stacked PR workflows before he convinced his team to give it a try. From there, interest spread organically. However, before Snowflake would make the leap more broadly, they needed proof of value, scalability, and a partner whose roadmap aligned with theirs.
This testimonial will cover how Snowflake adopted Graphite’s stacking workflow and AI code reviewer, Graphite Agent, across their eng teams to unify development and review cycles.
Stacked pull requests: Driving throughput for power users
Before broad adoption of Graphite, various teams at Snowflake procured their own developer tools in a piecemeal way, but sought a more unified approach that could scale across the organization. Snowflake inherently faced several challenges: inconsistent tooling across different teams created friction in handoffs and visibility, internal bandwidth to build and maintain a code review augmentation tool was more limited, and some engineers were skeptical about adopting a new workflow or command set.
Nathan Hunt, a leader on the developer productivity team, recalls, “As our team began looking for ways to solve bottlenecks in code reviews at scale, Graphite emerged as an obvious solution. Not only is there very little competition in this space, but Graphite's range of products expand across a number of parts of the software development lifecycle. To top it off, the pricing model is really fair, since you only pay for users who adopt it.
Stacked PRs quickly proved most valuable for Snowflake’s high-volume engineers. Adoption started out small, but over time adoption has grown to about 30% of Snowflake’s engineers, with power users leading the way. For these developers, stacked PRs are indispensable because they remove bottlenecks, allowing them to keep shipping without waiting on reviews. In turn, this makes the review process itself easier for reviewers.
Snowflake engineers that were early adopters of Graphite were shipping as many as 55% more PRs per engineer on average while outputting approximately 35% more lines of code per engineer.
AI-assisted code reviews with immediate value
Graphite’s AI reviewer added another layer of impact beyond stacked PRs. In practice, Snowflake engineers found it remarkably effective at surfacing subtle issues. Nathan explains, “Graphite Agent catches so many small logic bugs — it’s catching something meaningful with every third PR I write. That’s real value.”
By catching bugs early and reliably, Graphite Agent helps reduce cognitive load and risk on large merges, both for the author and reviewers, while reinforcing confidence in the tool’s output. Nathan adds, “Developers are skeptical of AI reviews when they’re noisy. But with Graphite, the comments are on par with human reviewers. I find myself thinking, ‘yeah, that’s right, I should fix that.’ It’s the kind of feedback you actually act on.”
That balance of high signal and low noise proved key to Snowflake’s adoption of Graphite Agent. Instead of being overwhelmed with suggestions, developers quickly saw it as a trusted collaborator, with engineers now accepting 55% of Graphite’s comments on their code. Over time, this level of engagement helped Graphite Agent become an integral part of Snowflake’s review process, particularly for high-volume contributors who benefit most from early detection of errors.
Why Graphite
For Snowflake, the decision to expand with Graphite went beyond feature offerings. There was also trust in the product’s roadmap and leadership team. “They’re [Graphite] communicative about where they’re headed. It feels like we’re in safe hands, with a team that’s out ahead of the challenges we’re thinking about too.”
Nathan sums it up: they didn’t feel the need to run a massive vendor bake-off. “Even if other competitors deliver comparable quality of code reviews in some dimensions, the value prop of Graphite remains stronger because of its integration into adjacent parts of the development experience,” he explains. Snowflake’s internally authored approach to AI code reviews is now complementary: Graphite handles broad, general tooling; Snowflake builds small, precise augmentations for their unique edge cases.
Looking ahead
As Snowflake and Graphite continue to collaborate, Snowflake is looking for ways to expand adoption beyond power users and encourage more of the team to embrace the stacked workflow. With trust in Graphite’s vision and ongoing feedback from Snowflake’s sophisticated engineering organization, the partnership is helping to scale code reviews, increase developer velocity, and reduce friction across the codebase. As Nathan puts it, “It’s exciting to feel like we’re solving these challenges together.”