Insights
Modernizing CI/CD with agents
How enterprise teams are reducing complexity while keeping engineers in control
Sai Gade | Global Head, AIDLC, Platform Engineering & SRE, UST
Enterprise CI/CD modernization is less about new platforms and more about how you work. Agents handle repetitive complexity. Engineers make the critical decisions. That’s how sustainable transformation happens at scale.
Sai Gade, Global Head, AIDLC, Platform Engineering & SRE, UST.
Enterprise CI/CD modernization is not a simple migration. It is an untangling of years of custom pipelines, interdependent systems, accumulated technical debt, and delivery commitments. At most organizations, these systems are working, which creates a deceptive stability. Everyone knows something is broken, but the status quo feels safer than the alternative. So teams continue managing layers of automation that nobody fully understands, maintaining Jenkins configurations that evolved through countless iterations, and accepting build times that frustrate engineers daily.
The question enterprise leaders are asking right now is not whether to modernize, but how to do it without grinding delivery to a halt. And more importantly, how to do it in a way that gives engineers better tools and less operational burden.
Agents are changing the equation for CI/CD modernization.
DIVIDER
Understanding the real cost of CI/CD complexity
Most platform engineering teams inherit CI/CD systems that were built incrementally, with each addition solving a specific problem at a specific moment. Jenkins rules are created for one team. Custom scripts are added for another. Workarounds accumulate. Over time, the system becomes a black box that requires specialized knowledge to operate.
The operational cost is real. Engineers spend time troubleshooting builds instead of building features. Platform teams maintain configurations nobody wants to touch. Migration efforts stall because the risk feels too high. And the business misses the opportunity cost: faster feedback loops, more reliable deployments, and the ability to scale development without proportionally scaling operations.
What CVS Health discovered is that this problem is not unique to them, and the solution requires more than a new platform. It requires a different approach to how you handle the repetitive, predictable work that consumes so much platform engineering energy.
DIVIDER
Agents as a modernization enabler
Agents are fundamentally different from automation scripts or workflow templates. An agent understands context. It can evaluate a situation, make decisions, and take action within defined guardrails. For CI/CD modernization, this matters tremendously.
Consider a typical build process: repository analysis, dependency resolution, test execution, security scanning, artifact generation, deployment staging. In traditional CI/CD systems, each step is explicitly configured, often with conditional logic, retries, and workarounds. If something changes, someone has to update the configuration. If edge cases emerge, someone has to add logic to handle them.
With agents, the system can learn patterns from your existing pipelines, understand what matters in your context, and handle the execution with less explicit configuration. More importantly, agents can reduce the noise that distracts engineers: they automate remediation of common failures, optimize retry logic, and surface only decisions that require human judgment.
But here is the critical part: agents are not about removing engineers from the process. They are about removing engineers from the repetitive parts of the process.
The CVS Health approach demonstrates this clearly. As teams migrated from Jenkins to GitHub Actions, agents handled the mechanical work of translating configurations, managing pipeline dependencies, and orchestrating the migration itself. Engineers stayed in control. They reviewed what agents proposed. They made decisions about what mattered. The result was a migration that moved faster because the tedium was removed, not because engineers were replaced.
DIVIDER
Planning a modernization that actually works
Enterprise CI/CD modernization succeeds when it acknowledges complexity rather than ignoring it. Here is what we have learned:
- First, map the real state of your systems. Not the documentation. The real state. Use agents to help with this discovery. They can analyze your existing pipelines, identify patterns, surface dependencies, and help you understand what is actually running, not what you think is running.
- Second, start with high-impact, lower-risk use cases. Migrate the straightforward pipelines first. Let the new system prove itself before you attempt complex migrations.
- Third, use agents to reduce the cognitive load of migration. Let them handle the repetitive pattern matching and configuration translation. This frees your platform team to focus on the judgment calls: prioritization, risk assessment, and integration strategy.
- Fourth, maintain oversight. Agents should surface what they are doing, propose changes transparently, and respect the guardrails you set. This is not about blind automation. This is about smart automation that keeps humans in the loop.
DIVIDER
Looking forward
The enterprises winning at modernization today are not the ones that migrated the fastest. They are the ones that addressed the underlying complexity while building a foundation for what comes next.
As agents become more capable, the organizations that learned how to work with them in CI/CD will be positioned to apply that approach to other parts of the software delivery lifecycle. Observability, infrastructure management, security operations. The pattern is the same: reduce repetitive work, keep humans in control, build for what is next.
CI/CD modernization is not about the new platform. It is about how you work. Agents make it possible to work differently.