AI Governance Framework
This is how we think about AI across everything we build. Short version: AI does the drafting, screening, and flagging. People decide. Here's what that means in practice.
Scope and oversight
Before we build any AI feature, we write down what it's allowed to do, what it's never allowed to do, and who has to sign off on it. That's decided at design time, not patched in later. Ownership of the outcome always sits with a named person on your team — never with the software, and never with us.
Risk management
We ask what happens if a feature gets it wrong, before it ships and again once it's in use — the closer a feature sits to money, compliance, or investor communications, the more scrutiny it gets. AI-generated answers have to trace back to a real source; if they can't, they're blocked rather than shown with a disclaimer. Where a product lets you tune how much AI can do on its own, we still hold the line at anything financial — no setting lets it skip human review there.
Lifecycle management
Shipping a feature isn't the end of the governance work. Every AI action is logged with who reviewed it and what was approved, so the record holds up if an auditor or regulator ever asks. Actions can be reversed — in products with an append-only ledger, nothing is overwritten; a correction is just a new entry, and the old one stays visible. Models that score or flag things get checked periodically for drift and bias, and a person can always override them.
Enablers
None of this works without the right foundations. We only use lawfully sourced data, and licensed data never gets redistributed. Where a product serves multiple clients on shared infrastructure, each client's data is walled off from every other's. New AI features ship behind a switch we can turn off instantly if something's wrong. And we revisit this framework as our products and the rules around them change.
This describes how AI is governed inside our own products. It doesn't replace the risk management or compliance obligations of the institutions who use them — those stay yours. AI may make mistakes. Always check if unsure. Questions about this framework: products@fundsview.tech.