Trust & Controls
A person stays in control of everything that matters.
Our systems touch customer communication, commercial data and — in mortgage — regulated work. This page explains exactly how that responsibility is handled: not as a disclaimer, but as the way the systems are built.
Agreed in writing before anything goes live.
Authority
What the system may do on its own, and where that authority stops. Delegation is written, bounded and revocable — never assumed.
Approval points
What waits for a person before it happens. In our current property system, no buyer receives a message until a person has read the thread and approved it.
Testing
What is tested against real cases before anything goes live. A system being tested is called a system being tested — not a launch.
Monitoring
What is watched every day once it's running, and by whom. Exceptions escalate to a person — ours first, then yours where judgement is needed.
Audit history
Every action logged, so any decision can be reconstructed months later: what the system saw, what it did, who approved it and when.
Data
Where your data lives and exactly who can reach it. In your systems wherever possible; anything held elsewhere is documented in writing first. Access is named, not shared.
What an approval actually looks like.
The system drafts from approved information, classifies the intent, pauses anything automatic, and queues the draft for a person. The operator approves, edits, or stands the message down. Every step is timestamped.
The boundary is fixed: AI drafts and classifies, rules decide what's allowed, people decide what matters — advice, credit, and anything a client will read.
09:41:09 campaign paused for this contact
09:41:10 draft created for operator review
09:44:31 approved by operator · logged
In regulated work, the line is firmer.
For mortgage brokerages: the broker remains responsible for the advice. The system handles the repeatable work around it.