---

Date & time

17:00

Teams are using AI to write and ship changes far faster than they can be vetted. Small live fixes to production pile up unrecorded, and by day 100 nobody can say with confidence what’s actually running, or why. GitOps closes that gap by providing a deployment and infrastructure layer that forces discipline, keeps one source of truth, and makes every change traceable.  Adopted by 77% of organisations, according to the CNCF, it has quickly become the safeguard on which AI-accelerated delivery depends.

But running GitOps isn’t the same as running it well. The credential model, rollback triggers, and repo structure all need tuning, especially as you scale from a pilot to widespread use.

Getting cleaner audits, faster recovery, and safer promotions means fixing the architecture: centralized or decentralized based on cluster context, automated rollbacks wired to observability metrics, and application code kept separate from deployment config so promotions stay reviewable. That upfront work pays for itself many times over.

AI agents are increasingly opening pull requests that get merged into your GitOps repo, which means a change can be fully tracked in Git and still be one that nobody actually reviewed or understands.

Done right, GitOps becomes what lets teams ship fast without breaking things along the way. This panel brings together engineering leaders who’ve built GitOps systems that hold up under real production pressure.

Key takeaways:

• Choose between a centralized and decentralized GitOps setup based on your team’s scale
• Decide what level of human review AI-generated infrastructure changes need before they’re allowed into your GitOps pipeline
• Wire automated rollback to real deployment metrics instead of manual dashboard-watching