In the first part of this series, I outlined the tensions emerging as engineering teams continue to use AI coding tools to accelerate delivery.
The State of DevOps Modernization Report from Harness found a clear correlation between heavy use of AI coding tools, faster delivery speed, and increased weakness in downstream processes.
For engineering leaders, these findings should not be taken as a warning to pull back on AI investment. But they are a reminder of the growing urgency with which we as an industry need to modernize DevOps pipelines and extend the use of AI further downstream.
Here are three priorities engineering leaders should focus on as they attempt to do so.
1. Standardize delivery pipelines
If every team builds and ships software differently, delivery pipelines quickly become fragile, and it’s easier for defects to slip through. Engineering teams come to rely on ad hoc workarounds, specialist knowledge, and manual effort to get code into production. That might have been manageable at the pace of traditional development, but it’s more difficult to sustain with AI increasing the flow of code into the pipeline.
As their teams continue to use AI coding tools, engineering leaders need to make it easier to do the right thing and difficult to do the wrong thing. Automated quality gates, rollback controls, security checks, and compliance standards should be built into every release by default, giving engineers a golden path to get their code into production.
These automated, standardized delivery pipelines help enforce a consistent level of quality, security, and resilience in software releases, so engineers can move faster within a system designed to support rather than hinder them.
2. Decouple deployment from release
Teams can’t innovate with confidence if release risk continues to increase with deployment frequency. Feature flags and progressive delivery practices are therefore becoming increasingly important as engineering leaders seek to address this imbalance.
Instead of treating every deployment as an all-or-nothing event, teams can stage incremental rollouts and test new releases with smaller user groups, so they can reverse course quickly if something goes wrong. This approach enables engineering teams to learn faster from their mistakes while simultaneously reducing the blast radius of failed releases.
For engineering leaders, feature flags and progressive delivery are among the most practical ways to enable higher delivery velocity without increasing the fragility of their pipelines.
3. Automate recovery, not just deployment
As engineers increasingly use AI coding tools, the speed of incident recovery becomes just as important as deployment speed. With teams pushing more changes through the pipeline, some failures will inevitably reach production. The question is whether teams can recover quickly from those failures, or if their process is defined by manual intervention, slow diagnosis, and individual heroics.
To ensure their recovery capabilities keep pace with the speed of AI-assisted coding, engineering leaders should focus on making rollback and remediation faster, more consistent, and more objective.
Service-level objectives (SLOs) that define a threshold for acceptable performance in production are central to this effort. By enforcing SLOs with observability data that provides real-time insights into application health, engineering teams can trigger autonomous workflows to roll back to a previous, stable version in the event of a failed release, without manually intervening.
This more intelligent approach reduces operational costs when things go wrong with AI-generated code, so engineering teams can innovate faster without fear of breaking things.
Turn AI speed gains into lasting value.
There’s growing evidence that AI is accelerating developers’ capacity to innovate at speed, but there’s little value in that capability by itself. Alongside the benefits, AI coding tools also create cost, complexity, and quality issues that delay delivery and increase risk.
The real goal should be to ensure that the faster pace of innovation being enabled by AI is both scalable and sustainable. Engineering leaders need to empower their teams to ship quickly while maintaining stability and reducing risk – all without adding extra weight to their shoulders.
DevOps modernization, underpinned by the three priorities I’ve outlined in this blog, is key to that objective.