The platform engineer's new job: Designing for agents, not just developers

Platform engineering has always been about reducing friction for developers through better tooling, golden paths, and internal developer platforms, all designed with one assumption: a human in the loop.

That assumption is breaking down. AI agents are now submitting PRs, triggering CI, and operating inside the same infrastructure platform teams built. But agents are not developers. They have no patience for ambiguous failure messages. They cannot intuit whether a test is flaky or broken. They do not escalate; they loop. A pipeline designed for human intuition becomes a trap for an agent.

In this talk, Eli Schleifer argues that the most important design question for platform teams in 2026 is not what tools to adopt, but who they are building for. He explains what changes when agents become primary consumers of a platform: what they need from CI feedback, why flake tolerance thresholds matter more than ever, and how merge infrastructure becomes the critical control plane for autonomous software delivery.

Platform engineers are uniquely positioned to own this transition. The question is whether they see it coming.