This resonates. We see the same pattern in financial modeling - the 'context layer' for financial models doesn't exist either. Models are flat spreadsheets with no entity relationships, no dependency graphs, no semantic structure. AI can't reason about =B47*C12 any more than it can resolve a customer across Stripe and Zendesk. We're working on this problem for finance specifically - structuring models as code with typed dependencies so AI has something meaningful to work with.