Most complex systems are already divided into sensible parts: product, technology, data, operations, and the people responsible for each. Individually, those parts can all be working reasonably well and the overall system can still fail.
I see this most often at the handoffs. A field means one thing in one system and something slightly different in another. A product decision creates an operational requirement nobody owns. A workflow makes sense to the team that designed it but not to the person expected to use it.
This is one of the most useful ways I use AI. Instead of asking it a narrow question, I give it several views of the same system: requirements, workflows, screenshots, data definitions, API documentation, business rules, and edge cases.
Then I ask it to look for inconsistencies, assumptions, and places where meaning changes as information moves from one part of the system to another.
AI does not replace the people who understand each part. It gives me a way to compare those perspectives at the same time and find the gaps between them.