Philosophy

Responsible AI Principles

The ten principles that govern how AI employees are designed, deployed and supervised.

Effective / reviewed: Reviewed 2026 — draft pending legal counsel sign-off

Human oversight

A person stays accountable for every consequential action. High-impact work routes to an approval queue rather than executing silently.

AI-generated content is labelled as such. Users can see which employee produced an output, when, and from what task.

Each AI employee has a named role, an owner and a record of the work it performed. Responsibility is never anonymous.

Data is processed for the purpose it was provided for. We minimise what we collect and keep customer knowledge scoped to the customer.

AI features inherit the platform's access controls. An AI employee can only reach data its role is permitted to reach.

The Brain view shows an employee's capabilities, knowledge sources and confidence, so output can be traced to its inputs.

You decide what to connect, what to train, what to approve, and what to delete. Automation can be paused at any time.

Performance is measured, reviewed and corrected. Training updates are logged to the activity feed.

Automation augments teams rather than concealing decisions. We avoid designs that pressure users into unreviewed action.

We do not design features intended to deceive, impersonate a real individual without consent, or manipulate people. Misuse is grounds for suspension.