Few services begin life with SRE support, so SRE needs a way to bring "the wisdom of production" to services old and new. Ch 32 ranks three approaches: 1. **Reactive onboarding queue** — evaluate existing services, negotiate fixes, and take them over in a prioritized queue. Common and reasonable in a fait-accompli environment, but it engages SRE *late*, after the service is already built and serving. 2. **Early consultation** — engage during design. Like a software bug, the earlier a reliability issue is found, the cheaper it is to fix; the service is more reliable "out of the gate" because SRE doesn't have to unwind suboptimal design later, and onboarding time drops. 3. **SRE-validated platform (best)** — give product development reliable, scalable, SRE-blessed infrastructure to build on. This short-circuits the stream of bespoke, individually-varying systems "arriving" at SRE's door, avoids whole classes of cognitive-load problems, and lets dev teams focus on application-layer innovation. These map onto the chapter's three engagement models — Simple PRR (reactive), Early Engagement (consultative), and Frameworks/Platform (designed-in). The trajectory runs from fixing reliability after the fact toward building it in by default.