Guest: Dimitri Vlachos (LinkedIn)
Company: Spacelift
Show: An Eye on AI
Topic: Infrastructure as Code

AI is changing infrastructure automation fast, but the real story isn’t about replacing Infrastructure as Code (IaC). It’s about giving teams two powerful deployment models that let them move at different speeds without losing control. At KubeCon in Atlanta, I sat down with Dimitri Vlachos, Chief Marketing Officer at Spacelift, to understand how intent-based automation is reshaping the workflow for DevOps, platform teams, and developers who want to move faster — safely.

Vlachos began with a clear message: the show floor conversations at KubeCon reflected a notable shift. Some teams are far into sophisticated automation practices, while others are only starting. But across the board, AI was on everyone’s mind — along with a healthy skepticism. Many vendors promise that AI can do everything, but most teams want to know where it genuinely helps and how it fits into existing processes. Vlachos sees this practical curiosity as a good sign. Teams want to experiment, but they want a safe environment to do it.

For years, Infrastructure as Code and GitOps have been the backbone of cloud automation. They bring rigor, process, predictability, and strong guardrails. Spacelift was built to accelerate this model and centralize it for enterprise teams. Vlachos still sees IaC as critical, especially for production workloads. But he also sees a gap: DevOps and platform teams spend too much time managing non-critical or short-lived environments. These environments matter, but they shouldn’t demand the same level of ceremony and code-heavy workflows as production deployments.

This is where Spacelift’s intent-based automation comes in. Instead of writing Terraform or HCL, teams can describe their desired infrastructure in natural language through their LLM of choice — ChatGPT, Claude, GitHub Copilot, or others. The LLM communicates with Spacelift’s MCP-based intent server, which applies policies, context, and guardrails so the deployment is safe and bounded. The most important piece, Vlachos emphasizes, is that the AI doesn’t get free rein. The policies that control IaC pipelines also apply to intent-based deployments.

Vlachos explains the model like this: IaC and GitOps remain the trusted path for high-stakes infrastructure. Intent-based automation becomes the fast lane for non-critical environments where iteration speed matters more than process overhead. Instead of writing and reviewing code, opening pull requests, waiting for plans to run, teams can deploy directly to the underlying cloud providers using AI — with safety layers in place.

The early access response surprised Vlachos and his team. They expected interest mostly from advanced IaC users wanting to offload repetitive work. They did see that, but they also saw something else: organizations with little or no IaC experience were using intent-based automation as an on-ramp. Some teams wanted to bypass the learning curve of HCL and get started with cloud deployments immediately. Others wanted the speed of intent-based workflows even for production, as long as the guardrails were there. Vlachos admits they didn’t expect people to push production workloads through intent, but strong policy support made it possible.

One unique element of Spacelift’s approach is how the company bridges both models. Everything deployed via intent can be promoted into an IaC pipeline. If a team iterates quickly on a non-critical environment and eventually wants to formalize it, Spacelift can automatically convert that infrastructure into code. This gives teams flexibility without locking them into one approach. Vlachos stresses they don’t see intent replacing IaC. Instead, most companies will run both paths depending on the workload.

Security is one of the biggest concerns teams have when introducing AI to infrastructure workflows.Vlachos addresses this directly. Spacelift provides context to the LLM so it understands the environment, and it enforces strict policies around what can be deployed, where it can be deployed, and who has access. AI cannot act outside these rules. He believes the industry should not avoid AI in infrastructure but instead understand how to safely use it in limited scopes before expanding adoption. Eventually, Vlachos expects agents — not just users prompting LLMs — to deploy infrastructure autonomously.

Beyond product announcements, Vlachos also talked about the growth of IaC Conf, the event Spacelift launched for the community. The first virtual edition saw over 3,000 registrations and more than 2,200 live attendees. The community wanted two things: deeper thematic spotlights and opportunities to meet in person. They delivered both by hosting a security-focused spotlight and their first in-person IaC Conf Connect during KubeCon. Despite travel challenges, the event drew strong attendance and a lively panel on AI in infrastructure, and even touched on how AI is shaping DevOps hiring.

Looking ahead, Vlachos sees AI as an accelerator, not a replacement for human expertise. The industry has always progressed through abstractions — tools that let teams focus on higher-level decisions rather than low-level mechanics. Intent-based automation is another abstraction layer, one that can free engineers to think about architecture, performance, reliability, and innovation instead of copy-pasting Terraform modules. But Vlachos is clear that creativity and understanding remain irreplaceable. AI should help teams move faster, not narrow their skills.

The conversation reinforced that the future of infrastructure isn’t a debate of IaC vs. AI. It’s about using the right tool for the right job, mixing rigor with speed, and giving engineers more freedom to focus on the work that matters. And with guardrails built at every layer, intent-based automation could become the next major abstraction shaping cloud operations.