AI in Infrastructure as Code: Justin O'Connor on Accelerating the Right Way

Josh Pollara • July 9, 2025

Justin O'Connor

Director at KPMG, Founder at Onward Platforms

Cloud Infrastructure & Platform Engineering Expert

About Justin

Justin brings a unique perspective to infrastructure as code, wearing dual hats as a Director at KPMG and Founder of Onward Platforms. His work spans from advising enterprise clients on cloud transformation to building developer tools that bridge the gap between traditional infrastructure management and modern AI-assisted workflows. With deep expertise in Terraform, cloud architecture, and platform engineering, Justin focuses on helping organizations balance innovation speed with operational excellence.

Current Focus: Developer-Centric Infrastructure Tools

Justin and his team at Onward Platforms are pioneering new approaches to infrastructure automation that keep humans in the loop while leveraging AI's capabilities. Their Developer Toolkit and Infracodebase projects exemplify this philosophy, using model context protocol (MCP) servers and intelligent agents to enhance, not replace, human decision-making in infrastructure management.

We will have more code, not fewer engineers. The need for expertise in the cloud might be greater than it ever has been.

Q1: Where have you seen AI genuinely transform infrastructure, and where's it mostly hype?

Language models and coding agents have significantly accelerated our teams' coding workflows. Tools like GitHub Copilot and Cursor, enhanced by model context protocol (MCP) servers, allow agents to autonomously create and push commits, raise pull requests, run safe Terraform commands, and access relevant documentation. These AI tools have been transformative for helping platform engineers quickly learn about cloud resources, troubleshoot runtime and deployment errors, and explore configuration settings. However, coding agents are typically better at standard programming languages like Python or Go. Because Terraform is declarative and tied closely to provider versions, agents can easily hallucinate plausible but incorrect configurations. They excel as "makers", writing more code, but often struggle as "maintainers," carefully refining existing infrastructure code.

The real hype is around the belief that language model-driven agents will soon replace platform engineers.

While AI is changing the nature and speed of the work, cloud infrastructure's complexity demands more, not less, human expertise. The true opportunity is designing AI-assisted systems that boost productivity without introducing additional risk, keeping humans involved and leveraging the deep expertise already present in the organization.

Q2: How are team roles evolving as AI becomes commonplace?

AI tools primarily offer speed and the capability to do tasks previously outside individual expertise. Yet, the "shift-left" movement, which transfers responsibilities from platform teams to developers, has largely failed in large enterprises due to the depth of cloud expertise required. It goes beyond just writing code and includes design patterns, operational requirements, business context, and security. The increased ease and speed of generating code has underscored the critical importance of clear business requirements, solid documentation, well-architected designs, early security integration, robust testing, and developer-friendly governance.

AI's potential lies in reducing friction and risk, making it possible to genuinely shift more work safely to developers.

Teams must prioritize clarity and alignment across the workflow, embedding requirements, architecture, and policy directly into the developer experience. When the right path is clear and easy, speed and quality improve simultaneously. AI, ultimately, should amplify each team's core strengths: developers build, platform teams enable, and security teams protect.

Q3: Fast-forward five years: how do you see us managing infrastructure day-to-day?

In five years, infrastructure management will become more automated and collaborative, though core fundamentals remain. Teams will still use declarative tools like Terraform (or its successor) to define infrastructure, but AI will handle significantly more of the heavy lifting. Day-to-day work will start with expressing intent through plain language or simple forms, defining requirements around performance, regions, and compliance. Specialized AI agents will generate the infrastructure code, handle integrations, and proactively flag issues. Engineers will focus on reviewing, refining, and approving generated outputs rather than manually writing code.

The greatest change will not be technological but organizational: increased transparency, better context sharing, and optimized balance between speed and control.

Workflows will integrate robust automated testing, security checks, and compliance enforcement by default, rather than as afterthoughts. Platform and security teams will spend more time designing the systems and policies guiding AI actions, reducing reactive ticket handling or configuration remediation.

Human review, however, will remain crucial, particularly in complex environments where small mistakes compound quickly.

Josh's Note

Justin's vision aligns closely with our experience at Terrateam: Automation and AI will transform infrastructure workflows, but humans must stay firmly in the loop to ensure quality and mitigate risk. Clarity, guardrails, and integrated policy frameworks are the keys to safely unlocking AI's potential.

Follow Justin's work:

LinkedIn - Connect for insights on enterprise cloud transformation and platform engineering
Onward Platforms - Learn about their innovative approaches to developer-centric infrastructure
Developer Toolkit - Explore tools that enhance infrastructure workflows with AI assistance
Infracodebase - Discover patterns and practices for modern infrastructure management

More in the Series

Petar Nikov on Avoiding AI pitfalls →

Kief Morris on infrastructure automation →

Get Started Read the Docs