How it usually works AI quietly scores and ranks your team, screens candidates in the dark, and your people data flows off to a model you cannot see.
What Capstan does Every entitlement and every access decision comes from a rule you can read, not a model guess.
Leave entitlements, access decisions, approval routing and generated letters are deterministic: the result comes from a rule you can audit, never a model guess. Payroll numbers are not ours to guess at either, for a different reason: Capstan computes no statutory pay at all. It compiles the inputs and hands them to your payroll partner, who computes the result, and Capstan holds no rate, slab or formula anywhere. No personal or sensitive record is ever sent to a third-party AI model or used to train one. One optional assistant, opt-in per workspace, answers a user's own question from aggregate figures they could already see and sends those to OpenAI, our one AI sub-processor, which does not train on the data. And no AI scores, ranks or screens your people, no attrition ratings, no sentiment analysis, no algorithmic candidate ranking, by documented product decision.