AI Usage Statement

AI Usage Statement

Version: 1.3 · Owner: Pocket PMO Ltd · Approval status: Approved · Review date: quarterly

Purpose

A short, customer-friendly summary of how Pocket PMO uses AI. This complements the AI Transparency and Responsible AI Policy.

What our AI does

• Drafts standups, executive summaries, plans, RAID themes, and content for reports.
• Suggests actions for the AI "project team" (Alex, Morgan, Jordan, Taylor, Sam).
• Provides a natural-language query surface.
• Transcribes and summarises meetings when invoked.
• Generates a stylised avatar from a photo a user chooses to upload, for their profile and AI team identities. No facial recognition, matching or identification, and the photo is not retained by the provider.
• Answers viewer questions on a published, read-only report strictly from that frozen report plus any private briefing note the publisher attached. The note is context supplied by the publisher, treated as data and never as instructions, is never displayed on the report, and is used for no other purpose.
• Monitors delivery signals continuously in the background and raises findings to your Agent Inbox.
• Drafts a concrete remedy for a finding when you ask for one, as a proposal you approve, edit or dismiss.
• Extracts editable risks, assumptions, issues and dependencies from notes pasted into the public RAID generator. This public tool cannot read or write customer workspaces and is protected by input, browser-use and IP-rate limits.

What our AI never does

• Change customer data without explicit user approval.
• Send emails or share content externally without explicit user approval.
• Act as an owner of tasks or work items.
• Make solely automated decisions with legal effects.
• Apply its own proposals. Background monitoring only observes; every change ends in a human approval, and applied changes are written to the audit log.

Where prompts and responses go

All AI requests are routed via edge functions to the Lovable AI Gateway. Model providers include Google (Gemini) and OpenAI. Prompts include only the specific project context needed for the task.

The public RAID generator sends only the notes the visitor deliberately pastes. They are processed transiently and are not saved by Pocket PMO as project content. Its output is probabilistic and must be reviewed before use.

Model providers, via the gateway, are contractually obliged not to train on customer content and to keep data transient. Provider attestations are held by Pocket PMO on request (owner-confirmation D-05).

Controls in your hands

• Turn AI features off at workspace level.
• Customise or disable AI Team personas.
• See AI credit usage in-product.
• Delete AI conversation history (owner-scoped RLS).
• Configure, in plain English, which delivery signals are monitored and what is drafted, at portfolio, client or project level.

Owner confirmation required

• D-05.

Document history