Governance model

Intelligence with defined authority

R.I.C.H.O. is built on the principle that AI systems must operate within declared boundaries, with human authority preserved at every consequential decision point.

Permission boundaries

Allowed
Analyse authorised information
Generate structured recommendations
Prepare internal plans and reports
Identify risks and evidence gaps
Create draft content
Route work for review
Approval Required
External communication
Public publication
Connected-system changes
Commercial commitments
Customer-facing deliverables
Material workflow execution
Restricted
Payments
Regulatory submissions
Legal representation
Use of private credentials
Unapproved external access
Irreversible actions without authority

Governance principles

Role-based AI assignment

Each tool deploys one named AI with a specific operational role. The AI cannot act outside its declared role.

Least-privilege permissions

Each AI is granted only the permissions required for its role. No AI has access to capabilities beyond its declared scope.

Human approval gates

Consequential actions are paused until a human explicitly approves them. No irreversible action occurs automatically.

Evidence status

Outputs distinguish between verified evidence, assumptions, unsupported claims and missing proof.

Audit trails

All activity is logged. Inputs, outputs, permission checks, approval decisions and blocked actions are recorded.

External-action restrictions

External actions — payments, publications, communications — are blocked unless explicitly authorised.

Data-handling rules

Submitted information is used only to generate the requested output. See the Privacy page for full details.

Safe failure

If the AI cannot produce a valid result, a clear error is returned. No partial output is presented as complete.

Human override

The system operator can pause, cancel or override any AI action at any point.

System status glossary

Interactive Demo

The interface and workflow are functional, but results may use scripted or illustrative logic.

Live AI Analysis

A secure AI backend generates a response from user-provided information.

Controlled Internal Operation

The AI can create approved internal records, drafts or workflow items.

Connected Operation

The system can use an authorised external integration within defined permissions.

Approval Required

The next action is paused until a human approves it.

External Action Blocked

The system cannot execute the requested external action.