Examples include finding relevant people at an account, researching possible professional connections and identifying a proposed introduction route. Specify the required output before selecting a provider.
A useful response contains stable identities, relationship type, source, evidence date, confidence and validation status. Keep inferred overlaps distinct from connector-confirmed relationships.
For influence research, ask why a person may matter to the decision. A graph edge does not by itself establish buying influence.
An assistant should only receive data the requesting user is authorised to access. Minimise sensitive fields, log relevant actions and separate reading from sending. A query result is not approval to introduce two people.
Confirm authentication, tenant boundaries, rate limits, errors, coverage and freshness. Evaluate results on known people before automating customer-facing actions.
Orbb's Network Graph and relationship-data direction is relevant to these needs. Ask Orbb for current API/MCP documentation, supported workflows and deployment requirements.