Platform engineering had a moment in 2022-2023. Every CTO presentation included a slide about internal developer platforms. The promise was clear: a self-service platform team builds golden paths, application teams move faster, ops burden shrinks.

By 2026, the actual pattern is less uniform than the original sales pitch. At some organisations, platform engineering has been the single best operational decision of the past five years. At others, it’s an underused internal product that the application teams ignore in favour of doing things their own way.

The pattern that distinguishes success from failure is depressingly simple. Platforms that started by talking to application teams about their actual pain points succeeded. Platforms that started by building what the platform team thought was elegant failed. This is the same lesson product management has been teaching for decades, just rediscovered in an internal context.

The other consistent factor is staffing. Platform teams that are properly resourced (typically 4-8 engineers at a 200-engineer company) tend to deliver. Platform teams that are 1-2 engineers asked to also handle on-call for the legacy infrastructure tend to underdeliver, then get blamed for the underdelivery.

What’s working in 2026: golden path templates for new services, paved-road CI/CD with security scanning baked in, observability defaults, environment provisioning. These have become genuine productivity multipliers at the orgs that did them well.

What’s still hard: platform discovery (no one can find the docs), platform versioning (every change breaks someone), and the political problem of legacy applications that the platform team is officially supposed to support but actually can’t help with much.

For IT leaders considering the platform investment, the honest read is that it works when treated as a multi-year product investment with proper staffing, real customer research, and senior leadership patience. It fails when treated as a cost-saving exercise to consolidate disparate ops teams. Same idea, completely different outcomes.