Operator Insights

Reported by others3

1. 

   James GarrettBackend developerTilt

   Lending company. Backend engineering on a monolithic codebase. Reviewers in the US and Australia.

   LeadDev, published Oct 8, 2025Watch the session (opens LeadDev in a new tab)

   Observation

   Tilt runs Auggie CLI as a review bot and reports faster code review across time zones.

   Takeaway

   Treat the review bot's rules as code. Keep them in the repository with named owners, and fix a wrong comment the way you would fix a bug.

   What happened

   A CTO-led AI enablement group trialed Augment on 12 seats, which grew to 24 as engineers asked for access. The team then built a pull request review bot that runs Auggie CLI in an Azure pipeline with rules the engineers wrote and own through a CODEOWNERS file. LeadDev's description of the session credits the bot with a drastic improvement in code review velocity across a wide range of time zones and with better first-time pull request approval rates. In his Tilt engineering blog post of Sep 11, 2025, Garrett credits the rules work: adoption only took hold once engineers invested in documentation, rules, and workflows.

   • Conference talk (LeadDev session produced with Augment Code)
   • Team outcome reported

   Deployment details

   Lending company. Backend engineering on a monolithic codebase. Reviewers in the US and Australia.

   • Conference talk (LeadDev session produced with Augment Code)
   • Team outcome reported

   Team

   Tilt backend engineering; trial grew from 12 to 24 seats

   Workload

   Pull request review bot on a monolithic codebase

   Product version

   Auggie CLI in an Azure pipeline, and the Augment IDE extension

   Measurement

   Code review velocity across time zones and first-time pull request approval rates, as LeadDev's session description states them, with no figure given

1. CollectorsOct 15, 2025LeadDevCollectors ran structured trials of AI coding tools and has begun shifting to Augment as its main tool.

   

   Dan Van TranCTOCollectors

   Collectibles grading company (PSA, PCGS). About 220 engineers, about 100 of them software engineers.

   LeadDev, published Oct 15, 2025Watch the session (opens LeadDev in a new tab)

   Observation

   Collectors ran structured trials of AI coding tools and has begun shifting to Augment as its main tool.

   Takeaway

   Run trials on production work with written criteria and a committee that owns them, and bring finance aggregate usage, keeping individual engineers' metrics out of it.

   What happened

   Collectors evaluates AI tools in three phases. It first checks pricing, documentation, integration, security, and vendor risk, then runs case studies in which real engineers aim for production-ready work, and finally scores each tool on a six-category matrix. An internal AI committee sets how tools are tested and secured. Staff engineer Alex calls Augment one of the clear winners among assistant tools. Alex credits its context awareness for rising adoption. The company, Alex reports, has begun shifting toward a single tool, primarily Augment. Van Tran aggregated usage data showing about 6% of the codebase written by AI tools to justify the spend to the CFO.

   • Conference talk (LeadDev session produced with Augment Code)
   • Organization metric reported

   Deployment details

   Collectibles grading company (PSA, PCGS). About 220 engineers, about 100 of them software engineers.

   • Conference talk (LeadDev session produced with Augment Code)
   • Organization metric reported

   Team

   Collectors engineering, about 220 engineers

   Workload

   Phased trials of AI coding tools on production work

   Product version

   Augment IDE extensions, evaluated against Cursor and Devin

   Measurement

   Share of the codebase written by AI tools (about 6%), adoption and engagement

2. RubrikJan 27, 2026Augment Code (YouTube)Rubrik's adoption rose once engineers saw Auggie carry a task to passing tests, then spread to triage and release work.

   

   Adam GeeVP of Engineering and founding engineerRubrik

   Security company. 12-year-old codebase across multiple product lines. Security-focused review of agent controls.

   Augment Code (YouTube), published Jan 27, 2026Watch the case study (opens Augment Code (YouTube) in a new tab)

   Observation

   Rubrik's adoption rose once engineers saw Auggie carry a task to passing tests, then spread to triage and release work.

   Takeaway

   Let engineers watch the agent finish a real task on your own codebase, and settle tool-call governance with security before adoption spreads.

   What happened

   Rubrik used Augment before Auggie with a handful of enthusiastic users. Gee reports that adoption rose to about 100% once engineers watched Auggie inspect the codebase, plan, edit, and rerun tests until they passed. Rubrik's security team also set controls on Auggie's tool calls. Teams now load task personas such as triage, review, and release coordination, with about 100 prompts contributed across the company.

   • Vendor case study (video published by Augment Code)
   • Adoption figure reported by the customer

   Deployment details

   Security company. 12-year-old codebase across multiple product lines. Security-focused review of agent controls.

   • Vendor case study (video published by Augment Code)
   • Adoption figure reported by the customer

   Team

   Rubrik engineering, with support staff also using it

   Workload

   Coding, triage, review, and release coordination on a large codebase

   Product version

   Auggie and the Augment IDE extensions

   Measurement

   Adoption of about 100%, as Gee states in Augment's video; ask Harness for a Rubrik reference call to see the usage data behind it