Twenty product teams each maintain their own CI/CD pipelines, databases, and monitoring. Design the fix, and tell me how you would know whether it worked.

1. Name the cost of the status quo: twenty near-identical implementations, twenty different failure modes, no shared upgrade path, and every new service starting from scratch. 2. Distinguish the models: DevOps automates delivery for one team, platform engineering builds a product other teams consume. 3. Build golden paths — opinionated, versioned templates for a pipeline, a database, and a monitoring stack — that produce a fully wired project. 4. Put them behind a self-service interface (portal or CLI) so a developer provisions in minutes without filing a ticket. 5. Treat the platform as a product with real product discipline: named users, documented contracts, versioning, deprecation policy, and a support channel. 6. Keep the escape hatch — teams with genuine needs outside the path must be able to opt out, or the platform becomes a bottleneck and gets circumvented. 7. Define success metrics up front: time from zero to a deployed service, share of services on the golden path, change failure rate, and time spent by product teams on infrastructure. 8. Watch the failure modes: a platform team that becomes a ticket queue, or golden paths so rigid that everyone builds around them.