# How data moves through Aiviance and AI A plain-language boundary for source data, AI processing, approvals and retained records. Aiviance separates where information comes from, where it is processed and who may act on it. Those are different questions, and this page follows the data through each stage. ## Source data enters a defined boundary Data arrives through a connected provider, an uploaded document or direct user entry. The source and owning entity remain attached to the record so that information from one person, family, business or trust is not treated as though it belongs to another. Connections follow their own consent and provider terms. Aiviance does not collect a user's online-banking password. ## Storage and AI processing are separate The primary Australian platform runs on AWS infrastructure in Sydney. AI requests are routed separately according to customer region, capability and data classification. For Australian customers, supported chat, reasoning, generation, vision and embedding workloads are routed to AWS Bedrock in Australia. If that eligible in-region route is unavailable, Aiviance does not silently fail over Australian content to a US model. A capability can instead be unavailable until an allowed route exists. The most sensitive data class is not sent to generative AI or large-language-model providers. Aiviance's current governed document-understanding register covers P1 and P2 data. The AI Use Disclosure separately permits an approved OCR processor to handle P0 data solely for document extraction where Aiviance enables such a flow; that permission must not be read as allowing P0 into a generative-model reconstruction step. Other AI inputs pass through the platform's privacy and policy controls before dispatch. ## AI prepares, people approve AI may classify, extract, reconcile or explain information. A model response does not by itself change a ledger, initiate a payment or approve a financial decision. The same entity scope, role, policy and confirmation controls used by the product remain in the path. ## Records remain traceable Provider, processing region, data classification and whether redaction was applied are recorded for AI invocations. Tool calls also produce a structured source record so a result can be traced internally to the data retrieval that produced it. That tool-level source record is currently best-effort, and a complete user-facing provenance view has not yet shipped. We therefore do not describe every AI answer as carrying a complete visible audit trail. ## What this page does not replace This overview is not a substitute for the Privacy Policy, Data Retention and Disposal Policy, AI Use Disclosure or a customer agreement. Those documents govern specific legal commitments and provider arrangements. ## Related policies - [Privacy Policy](/privacy) - [AI Use Disclosure](/legal/ai-use) - [Sub-processor register](/legal/sub-processors)