HEPO is intended to work alongside existing business systems. Agree which system owns each record and transaction. Integrations can bring evidence into HEPO and perform explicitly authorised outbound operations; they are not universally read-only. Read more.

Yes, where an outbound operation is implemented, configured and authorised. Separate incoming evidence from commands or updates sent to another system. Agree permitted operations, credentials, human gates, failure handling and reconciliation before enabling them. Read more.

Registered interface capabilities can be associated with process steps. Reuse depends on the data contract, action and interaction needed; a new requirement can still need configuration or implementation. Review both the flow and the user experience. Read more.

Background tasks and external integrations may use scoped execution credentials. Human decisions should retain their actor context, while the executing service and external identity remain distinguishable. Review permissions across each step rather than assuming the same credential is used everywhere. Read more.

Use the supported sandbox or simulation path for the process being evaluated. Define what is simulated, what uses test data and whether any external effects are disabled. Inspect the state and results before enabling the live operation; a successful rehearsal is not a guarantee of production behaviour. Read more.

HEPO describes its assistant around registered, typed capabilities and process operations. For the workflow being evaluated, inspect the actual tool or interface used and the permissions it carries. This is the basis for explaining the action, rather than treating a visual impression of a screen as an API contract. Read more.