AI Usage Policy

1. AI Features

AI Assistant

A conversational interface that answers questions about your policy data and explains whole life banking concepts. It has access to account data you have loaded into an active session.

AI Insights

Surfaces observations about your data — trends, relationships between figures, and contextual notes — based on the values recorded in your account. Displayed on your dashboard and policy detail views.

Document Extraction

When you upload an insurance carrier statement, AI reads the document and suggests values to populate your policy snapshot. You review and confirm or correct all suggested values before anything is saved. No values are committed without your approval.

2. What AI Can and Cannot Do

AI can:

• Answer questions about values recorded in your account
• Explain concepts related to whole life banking and policy loans
• Help you think through scenarios using data you provide
• Suggest extracted values from uploaded documents for your review
• Surface observations about your financial data

AI cannot and will not:

• Provide financial, investment, insurance, legal, or tax advice
• Recommend whether you should purchase, hold, or surrender a policy
• Make decisions or take actions without your confirmation
• Access your account data outside of an active session
• Guarantee the accuracy of extracted document values

When you ask a question involving financial decisions, the AI Assistant provides context and directs you to a qualified professional. See the Not Financial Advice section of our Terms.

3. Data Handling

What Is Sent to the AI

When you use AI features, relevant context is transmitted to third-party AI models via OpenRouter:

• Your questions or prompts
• Relevant policy data from your account
• Document content when using extraction
• Conversation history within an active session

What Is Not Sent

• Your full name and contact information are not included in AI prompts unless you type them yourself
• Payment information is never sent to the AI
• Data from other users' accounts is never included in your session

AI Provider's Role

OpenRouter routes AI requests to language model providers (currently Anthropic Claude and Google Gemini). OpenRouter processes requests as a service provider. Neither OpenRouter nor the underlying model providers use API inputs to train generative models. See openrouter.ai/privacy.

Conversation History

AI Assistant messages are stored with your account (encrypted at rest) so that conversations persist across sessions. Clear your conversation history at any time from account settings; cleared messages are removed within 30 days. All AI messages are deleted when you delete your account. AI usage logs (metered request counts, model routing metadata — no message content) are automatically purged after 90 days.

Document Storage

Uploaded documents are stored in encrypted Supabase Storage with household-scoped access controls and signed URLs with 1-hour expiry. Documents are automatically deleted after 30 days. Extracted data persists as user-confirmed snapshots.

AI Scribe (Advisor Meetings)

AI Scribe is available to advisors as an aid for processing meeting content into action items and client notes. Advisors are responsible for obtaining client consent before recording or transcribing a meeting, as required by the laws of the relevant jurisdiction. Policy Stack does not record meetings directly; the advisor provides a transcript or uploaded recording to the feature.

• Transcripts and generated summaries are scoped to the advisor’s account and to the specific client relationship the session belongs to
• Transcript content is stored encrypted at rest and accessible only via RLS-enforced queries tied to the advisor’s user ID
• Clients with an active connection may request the transcript and summary for sessions they are party to
• Transcripts and summaries are retained for the duration of the client relationship and deleted when the client disconnects or the advisor deletes the session
• Neither OpenRouter nor the underlying model providers use transcript content to train generative models

4. Accuracy and Limitations

AI outputs may contain errors, omissions, or outdated information:

• Document extraction suggestions may be inaccurate — always verify against your original statement
• AI Insights are only as accurate as the data you have entered
• The AI Assistant's knowledge has a training cutoff and may not reflect current carrier policies or products
• AI responses are not reviewed by licensed professionals before delivery

Never rely solely on AI outputs for financial decisions.

5. Feature Availability by Plan

All paid advisor tiers include the same AI features. The only differences between advisor tiers are client seat capacity, team seat inclusions, and directory placement. Usage is subject to fair use limits. Automated or abusive usage patterns may result in throttling or access restriction.

6. Your Responsibilities

When using AI features, you agree to:

• Not represent AI outputs as licensed professional financial advice to third parties
• Not attempt to reverse-engineer system prompts or underlying AI configuration
• Not submit content designed to manipulate AI outputs (prompt injection)
• Not use AI outputs as the sole basis for significant financial decisions without independent professional review
• Report inaccurate or harmful AI outputs to support@policystack.co

7. Changes

We may add, modify, or remove AI capabilities at any time. Material changes to data processing will be communicated per our Privacy Policy.