Fin says that over 93% of its pull requests (PRs) across its two main codebases are Agent-driven, and over 19% are auto-approved with no human reviewer in the loop. A controlled pilot of over 100 PRs through the AI approval pipeline produced zero reverts of AI-approved PRs and a 6–16x improvement in time-to-approval at the 75th percentile. In the first four weeks of broader rollout, 497 PRs went fully autonomous, with Claude writing the code and Fin's AI approval system reviewing, approving and shipping to production.
Post authors Kesha and Niamh Young write: "speed is not the enemy of safety. It’s a prerequisite for it. Accumulating code creates risk. Shipping small batches minimizes it."
The company's CTO, Darragh Curran, set a goal to double the productivity of the entire R&D organization within 12 months, and nine months later it was reached. Downtime from breaking code changes dropped 35% even as deployments doubled. PR review was one of the biggest bottlenecks: humans do not have the time to properly review the volume of AI-generated code, so either the queue backs up or reviewers start rubber-stamping.
The PR review Agent decomposes review into separate sub-jobs, each handled by an independent sub-Agent. One assesses the quality of the problem description, another checks whether the diff aligns with the stated intent, and others review for safety concerns, logical correctness, and best practices and known anti-patterns. The Agent also traces execution paths, following a change through the codebase. In testing on historical PRs, it flagged a one-line text copy change that contradicted an existing validation mechanism elsewhere in the codebase. The review is grounded in Fin-specific guidance that engineers build and refine, and engineers flag whether review comments were helpful, which sharpens that guidance. The Agent will not approve large PRs and requires them to be broken down. Any engineer can request a human review at any time, and the engineer who ships a change remains accountable for it.
AI-authored backend code had a revert rate of 0.53%, compared to 5.39% for human-authored code, and on the frontend the rates were 0.22% versus 2.00%. Every AI-approved PR is labelled, logged and queryable, with the review comments, approval decision, test results and merge event recorded. Fin engaged its auditors, Schellman, early to confirm that its automated review processes and the evidence they produce meet the requirements of compliance frameworks including SOC 2, HIPAA, ISO 27001, ISO 42001 and AIUC-1. In parallel, the team is working on an Agent that proactively diagnoses issues in production.