Conduct an owner-representative conversation

Core meaning

Capability purpose: Make selected owner knowledge and perspective conversationally available to others without exposing the owner's private Core or granting the representative unrestricted authority.
Meaningful outcome: A participant receives an attributable response from the configured representative, or the interaction is declined or escalated, with conversation and governance evidence preserved.
Boundaries — includes: representative identity, AI disclosure, audience and access state, dedicated knowledge scope, retrieval boundaries, conversation context, response provenance, refusal, escalation, session lifecycle, and audit evidence.
Boundaries — excludes: maintaining the agent definition, selecting the outward identity mode, populating the knowledge scope, granting publication/access rights, acting through the owner's undisclosed identity, and executing unrelated external actions.
Terms and concepts: An owner-representative conversation may map to schema.org/Conversation; participants map to schema.org/Person or relevant organization identities. An AI Twin is one possible product realization, not the generic capability itself.

Actor: An external participant and an owner-authorized representative agent.
Command / intent: Ask, discuss, or request information within the representative's declared purpose and audience scope.
Current state: The representative identity, disclosure posture, participant access, dedicated knowledge scope, conversation state, current policy, and escalation path can be determined.
Policies / invariants: AI nature and owner relationship are disclosed for this representative mode; only explicitly allocated outward knowledge is available; private Core and private assistant conversations are excluded; unsupported, sensitive, or unauthorized requests are refused or escalated; responses and source use are attributable.
Transition: Authenticate or admit the participant as required, assemble permitted context, generate or select a response, apply safety and authority checks, deliver or refuse it, and record the interaction.
Result: A delivered representative response, clarification request, refusal, escalation, or failed interaction with provenance.
Events / effects: May create conversation events, owner attention items, proposals, or follow-up work through separate capabilities.
Unknowns: Universal disclosure wording, response-liability treatment, escalation timing, retention, and participant-consent requirements are not established.

An outward representative must not inherit the owner's private Core or private-agent memory by default.
Representative conversation authority does not imply authority to transact, publish new material, or act through the owner's undisclosed identity.

Use a purpose-specific knowledge scope, visible AI disclosure, bounded audience, source-aware responses, and an accessible owner-escalation path.

Multi-turn consent, correction, abuse handling, and cross-channel continuity require implementation and governance profiles.

Statement	Semantic class	Evidence status	Source / note
Owner-representative AI conversation uses a dedicated outward knowledge scope, explicit identity/disclosure, controlled audience, and owner governance.	rule/invariant	sourced	J10–J12.