Computer Systems Design and Related Services2018Machine Learning (classification)Predictive AnalyticsB2C
Ubisoft

Ubisoft's AI Commit Assistant Catches 70% of Bugs Before Developers Commit Code

By training an AI assistant on ten years of game code, Ubisoft's R&D division built a tool that flags likely bugs before they enter the codebase — helping programmers make smarter decisions and planners build more accurate test schedules.

Bugs Caught Pre-Commit70%
4 min read

Background

In large AAA game development, bugs discovered late in the testing phase are expensive — they delay launches, consume QA resources, and erode team morale. Ubisoft sought to shift bug discovery upstream, closer to the point of code authoring, so that quality problems could be addressed before they compounded.

What Was Implemented

  • Ubisoft R&D, in partnership with Canadian universities, trained an AI assistant on ten years of internal game code
  • The "Commit Assistant" analyzes code at the point of commit and returns a probability score for bug likelihood
  • The tool (later evolved into "Clever Commit") was applied to 12 internal systems
  • Programmers receive real-time statistical guidance without requiring a separate QA cycle

Results

The Commit Assistant prototype detected 70% of bugs before they were committed. The later Clever Commit iteration achieved 79% accuracy in detecting bug-prone code and provided fix recommendations in 66.7% of cases . These figures come from 2018 press coverage; no formal peer-reviewed study was located.

Lessons

  • Training AI on years of proprietary code history creates a bug-prediction model grounded in the company's actual failure patterns
  • Early bug detection shifts quality costs upstream and reduces last-minute testing crunch
  • Partnering R&D teams with academic institutions accelerates prototype-to-production timelines

Ready to implement AI in your commerce operations?

McFadyen Digital helps teams move from case study to live implementation.

Talk to an expert →