BankingNewsAI Daily Brief ·
EXANTE's Codey reaches a 75% success rate while humans stay in control
Banking AI
Financial institutions & fintech technology
EXANTE's Codey hits 75% success, but humans keep control
For the CIO of a regional bank piloting coding agents
One in four Codey tasks failed, and every release still waits on a human signature. That is the honest ceiling for agent autonomy in your change-management policy, whatever the demo shows.
Codey is EXANTE's in-house coding system: six AI agents, from researcher to verifier, that take a Jira ticket and return a merge request with a preview link, typically in about 30 minutes. Of 250 tasks, 75% succeeded. The simple model is a fast junior developer who never sleeps. It is imperfect. Failures cluster where context is missing, where changes span linked repositories, or on a first task in an unfamiliar stack, which fits plenty of bank code. The exception that matters is the cap: after three revision rounds a task is marked "AI failed", and humans keep acceptance and release. So why does a faster coder add work? Because review shifts to people. Codey handles 7% of its teams' work. The sign-off, not the typing, sets the speed.
→ Action
Change management: Require named human approval and a capped revision count on every agent-authored merge request before production release.
Full story →Read article → from FinTech Global
Get this in your inbox every morning
Free · No spam · Unsubscribe anytime