Answer
The architecture is typically hybrid: core transactional workloads, the parts that need to be fast, auditable, and always available, stay on IBM i where they already run reliably, while AI models run in IBM Cloud and data moves between the two environments through connections built for that purpose. For regulated industries, watsonx.governance adds monitoring and audit trail capability on top of that connection, which matters for banking, healthcare, and insurance buyers who need to show a regulator exactly what the AI accessed and recommended.
Buyers evaluating vendors should ask specifically how the AI tool connects to IBM i, whether it requires read-only or read-write access, and what governance or audit logging is included, since those answers matter more than which model the vendor is using.