Revolut Cuts Coding Time by 30-60% Using OpenAI Models and Codex
·2026.11.16 18:30
Key point
The bank uses AI agents to automate financial crime screening and accelerate software development while maintaining strict human oversight.
Details
- Revolut reports a 30-60% reduction in coding time from task start to feature merge by using OpenAI models and Codex to delegate implementation tasks to multiple agents.
- A FinCrime agent powered by OpenAI models now handles transaction screening and initial investigations, allowing the bank to monitor more activity without increasing investigator headcount proportionally.
- Engineers use Codex to build internal tools, such as a code-review service, by assigning specific tasks to agents while focusing on system design and solution planning.
- Security teams saved 1-2 days on a substantial code-review task and used AI to instantly trace phishing connections, though humans remain responsible for final approval and accountability.
- Revolut converted seven years of design documentation into AI-readable context in one day to improve accessibility of organizational knowledge for designers.
- The company maintains a systematic framework where all AI-built solutions undergo the same standards for development, A/B testing, and business metric prediction.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.