If a Digital Twin uses a Digital Representative for customer support, how do you manage escalation to human agents when the AI can't answer?

Establish an escalation workflow with clear triggers (uncertainty thresholds, unsupported intents, or user requests for human assistance). When triggered, the system should transfer the current conversation context to a human agent, preserve any visible proofs or hub facts referenced, and provide the customer with expected response times and escalation tracking. The handoff should respect privacy boundaries by not exposing internal procedures, and the interaction should be logged in the Change Log for auditability. This keeps users supported while maintaining trust and governance around AI capabilities.