Systems engineers have an answer to whether AI can help integrate their models across companies: yes, but only if a human checks the work.
A paper posted to arXiv on September 30, 2026, titled "Cross-Organizational SysML Model Integration: A Survey of Challenges and AI-Supported Tasks," surveyed 29 stakeholders who work with SysML-based Model-Based Systems Engineering (MBSE) across organizational boundaries. Respondents rated eight predefined categories of integration challenges and six types of AI-supported tasks on five-point scales. The results show model integration is less a single technical snag than a multi-dimensional alignment problem spanning semantics, behavior, traceability, and how systems exchange data. How respondents ranked those challenges varied by their organizational role and how often they handled cross-org integration work.
MBSE tools like SysML coordinate complex builds (aircraft, vehicles, defense systems) across multiple companies and suppliers, which is exactly where integration tends to break down. The survey found AI rated highly useful for analysis tasks, such as parsing semantic structure and flagging inconsistencies between models, but respondents overwhelmingly preferred a human in the loop with mandatory verification rather than AI making integration calls unsupervised.
That preference echoes what has happened with AI coding assistants, which have settled into a review-and-suggest role rather than an autonomous one. Engineers, per this survey, want the same arrangement for their system models: let AI flag the problems, but keep a person signing off on the fix.