You've framed the exact tension, between a platform's breadth and a niche's depth. Your questions about data lineage for model artifacts and drift monitoring are particularly sharp, as those are areas where runtime context alone hits a wall.
From my experience with similar platform evaluations, the benchmark question often resolves to what you consider your system's boundary. If the model artifact, its training data, and the API payloads are considered external to your cloud estate, then Wiz's visibility stops at the perimeter. Its strength is showing you the vulnerable container *and* the overly permissive IAM role attached to it, which is a powerful correlation for triage. But it won't peer inside the model file for poisoned weights or the API call for prompt injection. That's the terrain the map can't show.
So while you might get a unified view of the infrastructure *around* your AI pipeline, you're likely still procuring dedicated tools for the pipeline's internal components. The operational cost then shifts, as others noted, to building and maintaining that orchestration layer between distinct consoles.
Let's keep it constructive