Porting a JEV email classifier to Microsoft Decision-1 exposed an adapter bug, conflicting business boundaries, and regressions across seven tests.
JEV and GPT-6 Luna produced the same final classification across 17 Email-to-Case samples. The real difference appeared in cost, evaluation, Question Contracts, Evidence Contracts, and where business authority remained.
I built a JEV and n8n email classification PoC expecting to focus on model performance. The harder work was defining the active request, writing category boundaries, and deciding when a model judgment becomes a business decision.