Capstan’s answer, in public
No AI scores, ranks or screens your people, and that is architectural rather than a promise: the agent capability registry is pinned in the database so it cannot hold a capability that evaluates a person, nothing of the kind is built, and the analytics engine can only ever narrow who sees what, with small groups suppressed. The useful AI is off until a workspace turns it on and is bounded by construction: an assistant answers from what the caller may already see, and an agent needs a named accountable human and an explicit human confirmation before it changes any record, can move no money, and can evaluate nobody. The one aggregate assistant that reaches out sends only non-personal figures, such as headcount by department, to our single AI sub-processor, OpenAI, which does not train on the data. No per-person or sensitive record is ever sent to an outside model or used to train one.