AI is Transforming Infrastructure as Code, But Humans Still …

Infrastructure as Code (IaC) has grown from a developer productivity hack into a cornerstone of modern cloud operations. Today, big tech firms rely on it to scale across the globe.

Netflix, for example, uses IaC to power everything from continuous delivery with Spinnaker to container orchestration with Titus, ensuring consistent deployments across its vast infrastructure. Similarly, Shopify’s Conversations team applies version control, testing and auditability to infrastructure changes that were once dependent on manual graphical user interface clicks.

This evolution is now entering a new phase with artificial intelligence. From template generation to compliance checks, AI is making IaC faster, smarter and more responsive. Yet, AIM heard from two engineers, Lukas Nießen, software architect at ista, and Utkarsh Kanwat, engineer at ANZ, that human oversight remains the critical safeguard in the next stage of evolution.

AI: From Acceleration to Dependence

For Nießen, AI tooling acts like an accelerator. “Similarly to how you can use LLMs to accelerate ‘normal coding’, it accelerates IaC,” he told AIM.  

He highlighted AI’s use case for developers as something that acts as “template authoring and search on steroids”. Yet, he warned that speed must not come at the cost of caution.

“One needs to be double careful with generating infrastructure code with AI, of course,” he added. Without strong checks, automation can amplify mistakes.

Kanwat shared that his team has taken it a step further by building a coding agent specialised in Terraform projects. “It analyses the project and makes surgical edits,” he explained, pointing to Cursor and Copilot as other valuable aids. 

Debugging, he noted, often involves extracting logs, feeding them into AI systems and identifying root causes quickly. However, he insists that accountability lies with the engineer.

“You still need to fully understand what AI is actually generating and how everything functions together,” Kanwat told AIM.

Complexity Doesn’t Disappear

AI may accelerate workflows, but it doesn’t erase complexity. Netflix’s ability to scale globally, along with case studies of other big companies, demonstrates the significant orchestration required under the hood. 

Nießen believes the current multi-cloud shift stems from a need for redundancy and independence. “This shift favours tools with stronger abstraction capabilities like Pulumi or Crossplane. However, Terraform seems to remain the most commonly used one, which is partially because of its mature provider ecosystem,” he said.

Kanwat observes a similar trend in enterprises chasing reliability. Multi-cloud strategies offer freedom from vendor lock-in. “Tools like Terraform and Pulumi definitely win because they avoid vendor lock-in, so you’re not stuck with AWS CloudFormation or Azure ARM templates. But the complexity trade-off is still very real,“ he explained.

For smaller teams, managing multiple pipelines and configurations can outweigh the benefits, a reminder that automation often amplifies scale challenges rather than erasing them.

Security and Governance: Guardrails Under Pressure

Even companies with advanced IaC pipelines face challenges with governance. Airbnb and Shopify, for instance, rely on Terraform to enforce versioning and auditability, but Nießen argues that evolving governance rules make compliance more difficult every year. While AI might eventually assist by catching policy gaps in real time, it is still in the early stages of development.

Kanwat identifies policy enforcement at scale as the true bottleneck. Teams struggle to apply consistent rules across cloud providers. Worse, policies that are too rigid can block developer productivity altogether. “The productivity killer is when policies are too rigid,” he noted. Automation can check policies, but alignment still requires careful calibration.

GitOps and Automation

The GitOps model shows where automation excels, by providing structured support for human-driven processes. 

For Nießen, it is a “must-have” for serious IaC, as it prevents drift and enables safe rollbacks. Kanwat praised Git’s built-in access controls and review processes, but pointed out the limitations. “Many IaC modules lack functionality, so you end up doing ClickOps anyway,” he said. 

Even at companies like Shopify, where Terraform enforces rigorous workflows, real-world incidents can still demand manual fixes.

From Netflix’s global deployments to Shopify’s Terraform-driven engineering culture, IaC is already indispensable for operating at scale. AI promises to supercharge this foundation by accelerating authoring, debugging and compliance checks. Yet, Nießen and Kanwat’s experiences highlight the same lesson: automation cannot replace judgment.

As enterprises embrace AI-driven IaC, success will depend less on eliminating human involvement and more on striking the right balance, allowing AI to handle the heavy lifting, while engineers maintain control of safety, governance and design.

The post AI is Transforming Infrastructure as Code, But Humans Still Hold the Reins appeared first on Analytics India Magazine.

Generated by RSStT. The copyright belongs to the original author.

Source