platform-org-design-advisor

The practitioner inflection point is roughly 100–150 developers, earlier for weak engineering cultures. Below ~50 developers, lay foundations without a standing team; use 10–20% of engineering capacity as the sizing heuristic once past the threshold.

A dedicated Head of Platform Engineering is the most common pattern at 32.9% of orgs, followed by VP Engineering at 21.1% and CTO at 14.6%. Reporting into Infra/Ops correlates with ticket-ops relapse, and any line more than two hops from a technology executive is a flag.

Default to platform-team ownership with named interfaces to data science, matching the 36.7% industry pattern. Escalate to a dedicated AI-platform team only at scale, such as multiple GPU-backed products, to avoid recreating the DevOps-silo mistake.

Human roles progress from Executor to Validator to Orchestrator to Constraint-setter across levels L0–L4. Each shift moves functions from doing to verifying to designing verification, requiring staffed review capacity at L2, dispatch ownership and eval engineering at L3, and governance-as-product at L4.

Benchmarks derive from 2025 industry surveys of roughly 500 and 240 respondents and are directional priors, not targets. Findings based on partial connector visibility or self-reported maturity are marked low-confidence rather than presented as certain.