AI Transparency Policy

AI powers the heart of Panzoe: assistant conversations, Panzoe Daily briefings, the AI Inbox triage and reply drafts, calendar insights, feedback analysis and cross-area recommendations. Wherever content is AI-generated, it is presented inside AI features — Panzoe does not disguise AI output as human-written communication.

Assistants operate on your workspace's shared AI Brain: the content you add plus summaries synced from integrations you chose to connect. Access is permission-filtered — workspace admins can block specific assistants or restrict integrations to specific areas, and those restrictions are enforced on every AI request, including when assistants consult each other.

Panzoe Memory & Context Engine — Panzoe may include a handful of your own recent memories and matching snippets from your own research, documents, conversations and records in an AI request to make the answer more useful. This retrieval only ever searches your own data within your workspace — never another user's. Memory can be reviewed, deleted or switched off in Settings → Privacy & data. Research uses your private Panzoe data only when you switch "Include my data" on, and those private-sourced findings are labelled as coming from your data.

Panzoe sends your prompt plus relevant, permission-filtered context to established AI model providers — principally OpenAI (text, voice transcription, speech and video generation) and Google (Gemini) — through a managed, secured inference gateway (Emergent) under contractual restrictions. We do not permit these providers to use your content to train their foundation models. Some features use AI 'tools' — structured actions such as searching your inbox — which are logged and, for any write action, require your explicit confirmation.

Delegated jobs — when you delegate an objective to the Panzoe Chief of Staff, Panzoe plans and runs the steps (research, writing, voice) using the same providers and rules; any step that sends content outside Panzoe waits for your approval first.

AI can be wrong. It may produce inaccurate, incomplete, outdated or fabricated information ('hallucinations'), misread context, or reflect biases in its training data. Panzoe reduces these risks by grounding responses in your real data, but cannot eliminate them. Always verify important information before acting on it.

AI output is informational assistance, not legal, medical, financial, tax or other professional advice. For decisions with significant consequences, consult a qualified professional. You remain responsible for the decisions you make and the actions you approve.

Connecting integrations is always optional and reversible. You can disconnect services, purge synced data, restrict assistants, view and delete individual memories, clear all memory, turn Panzoe Memory off entirely, and delete your account. All of these controls live in Settings → Privacy & data. AI features degrade gracefully — Panzoe never requires you to connect anything.