# How Ripple Treasury automated bug triage and tripled resolution speed with Devin ![Vimeo](https://devin.ai/_next/image?url=%2F_next%2Fstatic%2Fimmutable%2Fmedia%2Fcover.1z27me3uh8wem.webp\&w=3840\&q=75) 75% Reduction in untriaged bug backlog 3x More bugs fixed per day 70 Teammates outside engineering using Devin to answer technical questions ## About the company [Ripple Treasury](https://ripple.com/) is Ripple's treasury management platform, helping CFOs and finance teams manage cash, payments, liquidity, and financial risk across traditional and digital assets. Formerly GTreasury, it was acquired by Ripple in 2025. Industry: Financial Services, Technology**[Visit site](https://ripple.com/)** Ripple Treasury is Ripple's treasury management platform, helping CFOs and finance teams manage cash, payments, liquidity, and financial risk across traditional and digital assets. Formerly GTreasury, it was acquired by Ripple in 2025 and now operates as Ripple Treasury. Jeremy Tuason leads the team that builds the financial instruments side of the product, including deal management, portfolio and exposure management, and accounting reports. Like any large enterprise software platform, the team receives a steady flow of customer reports that need investigation and triage. As customer growth drove that volume up, the manual refinement process consumed more and more senior engineering time, and in the last year the number of issues awaiting technical triage grew faster than the team could work through them. Before an engineer could start on a ticket, engineering leadership had to investigate it, identify the root cause, document steps to reproduce, and turn it into something the team could estimate and refine. Leadership met twice a week for dedicated triage sessions, which cut into the time senior engineers could spend on roadmap work. ## Building an automated bug triage system Ripple Treasury started using Devin in 2025 as a pair programmer. Engineers used it to answer questions about the codebase that previously took hours, or sometimes days, of manual investigation. With the right context, Devin returned useful analysis in minutes. In February 2026, the team used Devin playbooks to build an automated bug-triage system called Guardians 2.0. Rather than give a single Devin session the entire job, the team modeled the workflow on the engineering roles in its real refinement process. Four Devin agents divide responsibility for investigating and challenging each bug: - **Catalina**, the orchestrator, enforces the team's Definition of Ready and validates the overall analysis. - **Torin**, the investigating engineer, analyzes the codebase, identifies reproducible failure scenarios, and develops an evidence-based root-cause analysis. - **Gilfoyle**, the tech lead and devil's advocate, challenges assumptions and rejects conclusions or changes that aren't sufficiently supported. - **Grace**, the SDET, looks for edge cases and regressions and validates the testing strategy. For every customer-reported bug, the workflow produces the artifacts engineers need before refinement: a technical analysis, steps to reproduce, and a development plan. The team reviews the output before a ticket moves forward, but senior engineers no longer perform the investigation from scratch. The untriaged backlog fell by 75%. The team went from fixing roughly 0.5 bugs per day to 1.5, a 3x increase, and eliminated its twice-weekly triage meetings, giving its most senior engineers more time for roadmap work. ## Reproducing and testing bugs in a real Windows environment Even after Guardians 2.0 investigated a bug, engineers still had to reproduce it, set up the right environment and test data, implement the change, and verify that the fix worked in the running application. In August 2026, the team began automating that work too. They added a fifth agent, Woz, running in Devin's Windows environment. Devin works inside a real Windows environment, building and launching the application, interacting with it in a browser, and testing its own changes. For Ripple Treasury, Woz can reproduce customer-reported behavior against real test data, implement a fix, and verify the result in the running application. Ripple Treasury configured the environment once, with its Windows development stack and dependencies ready to go. Each session starts from a ready snapshot, and the only manual setup step is approving VPN MFA once per session. Woz starts with the technical analysis from the upstream agents, then takes the bug through the rest of the loop: reproduce the issue, implement the fix, run the application, verify the fix, and capture the evidence. For every bug, it delivers a pre-fix reproduction video, the pull request, a post-fix verification video, and testing screenshots. Engineers review the change alongside evidence of the original bug and the working fix instead of starting from a proposed code change and reproducing it themselves. ## Giving customer-facing teams faster access to technical answers Support, implementation, and solutions teams regularly need to answer detailed questions about how a module behaves, whether a particular workflow is supported, or why a customer is seeing a particular result. Historically, many of those questions eventually reached an engineer. Now those teams use Devin to investigate technical questions themselves, with access governed by Ripple Treasury's existing controls. From March to August, the number of technical questions answered with Devin grew from 447 to 822 per month. Today, over 70 team members use Devin for this work. Across 4,600+ technical questions to date, Ripple Treasury estimates that Devin has saved roughly 2,300 hours of engineering investigation, assuming 30 minutes of engineering work per question. That is equivalent to over 280 eight-hour engineering days. Engineers stay focused on building while customer-facing teams investigate the codebase themselves. ## What's next The next step is to integrate Grace directly after Woz, so that once a bug has been reproduced and a fix verified, Devin automatically generates the corresponding test cases and attaches them to the ticket. Engineers could then review the technical analysis, reproduction, code change, verification evidence, and test coverage together. From there, Ripple Treasury plans to continue reducing the remaining triage backlog, measure changes in lead time and first-pass review acceptance, and extend the same environment blueprint to other codebases and engineering workflows. The broader goal is to move more of the customer-issue workflow away from manual investigation and toward work that arrives ready for engineers to review. As the Ripple Treasury team describes the shift internally, instead of spending most of the sprint figuring things out, engineering can spend more of the sprint delivering outcomes.