Alright, let’s face it: the way we build and ship software today is… kinda broken. We’ve got a gazillion tools, cloud-native everything, YAMLs coming out of our ears, and enough complexity to make even the most seasoned engineer cry into their terminal. So what did we do to deal with it? We created Internal Developer Platforms (IDPs). Cool, right? They were supposed to simplify things, reduce cognitive load, abstract away the madness. And for a while, they did. But let’s not kid ourselves: even IDPs aren’t immune to entropy. Give it enough time, and your precious platform becomes yet another spaghetti monster of tools, interfaces, and tribal knowledge.
Here’s the truth: we need new interfaces. And that interface is AI. Not just any AI, we’re talking generative AI, the kind that can read, write, and (if you squint hard enough) think like an engineer. This is where the real platform revolution starts. The next evolution of developer experience isn’t just about slick dashboards or opinionated CLIs. It’s about having an AI sidekick that actually gets it. An AI that understands your stack, your code, your configs, your weird internal acronyms. But here’s the catch: AI without context is just a fancy autocomplete. You throw it a question, it throws back something plausible-sounding but useless. That’s not intelligence: that’s improv comedy.
So how do we fix that? Context. Real, rich, structured context. And guess what gives us that? A platform built with an everything-as-code mindset. You know, manifests, configuration files, version-controlled infra, service definitions and all the stuff we already have lying around. When you wire all that into your platform and let the AI feast on it, magic happens. Suddenly, it’s not just generating fluff, it’s generating relevant, accurate, actionable insights. Because it knows how your world works. It’s not hallucinating answers anymore, it’s navigating a well-lit map.
Let me throw a few examples at you to paint the picture. These are some new features we’re actively working on and that we showcased just a few weeks ago at KubeCon EU, early-stage ideas that hint at the future we’re building, where AI becomes a first-class citizen in platform engineering.
First up: onboarding. Everyone hates onboarding. It’s slow, clunky, and full of awkward questions like “Where do I find the API key for staging again?” Now imagine you’ve got a platform with solid documentation, all indexed and searchable by a generative AI assistant. But not just any assistant, one that understands the docs because it’s been trained on them. A RAG system that can answer not just with confidence, but with context. New devs show up, start asking questions, and boom, they get answers tailored to how your platform works, not some generic Stack Overflow nonsense. It’s like having your smartest staff engineer available 24/7, minus the burnout.
Next: discoverability. Good platforms are composable. You’ve got a catalog of reusable components, services, templates: all the building blocks for modern applications. Problem is, finding the right piece at the right time is often a nightmare. You need to know what exists, what it does, how it connects. Enter generative AI again. It knows the catalog. It knows the relationships. It can suggest components, explain how they work, even draft the glue code to wire them up. It’s like having an architect whisper in your ear, “Hey, you don’t need to reinvent that, we already have a module for that exact use case.”
And then there’s troubleshooting. Oh boy. Distributed systems are amazing until they break, and then they’re a giant pain. Logs everywhere, metrics in five different dashboards, alerts screaming in Slack. Who has time to grep through all that noise? But your platform already collects all this data: logs, events, traces, metrics. What if your AI could look at all that, correlate it, and give you a summary like, “Looks like the payment service started failing right after the last deployment, and it’s throwing timeouts talking to Redis. Might be a connection pool issue.” That’s not just helpful, that’s game-changing. That’s the kind of assistant you actually want on your team.
Now, you might be thinking, “Sounds great, but it also sounds like a lot of work.” And yeah, getting there takes effort. You need clean, standardized inputs. You need to break down the silos. You need to think of your platform not just as a delivery mechanism, but as a context engine for AI. But here’s the thing: you’re already halfway there. If you’re doing GitOps, if you’re treating infra as code, if you’re building service catalogs and writing documentation you’ve already got the ingredients. Now it’s about connecting the dots and letting AI do what it does best: synthesize, assist, and accelerate.
This is the future of platform engineering. Not just managing tools, but designing systems that can teach and adapt. Not just building interfaces for humans, but also for machines that can partner with us. We don’t need another dashboard. We need intelligence. And not artificial intelligence in the buzzword sense but real, contextual, helpful intelligence that makes our lives as engineers easier.
So yeah, maybe it’s time to stop thinking of AI as some shiny add-on and start treating it as the next layer of the platform. The interface that finally tames the complexity we’ve all been drowning in. The co-pilot that doesn’t just autocomplete your thoughts, but actually understands your mission. If that doesn’t revolutionize software delivery, I don’t know what will.