Copilots in Saolix products

AI agents that advise, never decide.

Where Saolix products use AI, it takes the form of an AI agent that helps people work faster. No agent can make the decision. In Eagle, that is enforced in the product, not left to policy.

Catalogue

The AI agents, and where they stand.

In Saolix Eagle · optional

Raia

AI agent for AML alert triage

Through Eagle's triage API, suggests a label for a screening alert, likely noise or needing review, with a confidence and a reason. It runs on a locally hosted open model. A guard in Eagle overrides any "noise" label on a real hit, and Raia can never close an alert.

Suggests a labelLocal modelCan be switched off
In Saolix F2X · concept

Agentic investigation copilot

Fraud & risk

Designed to link related activity across customers, accounts and devices and explain why a pattern looks like fraud. Governed rules decide whether to hold, block or allow.

In development
In Saolix Shield · concept

Agentic security copilot

Cyber threat response

Designed to correlate security events into a likely attack path and suggest a response. The response runs only under approved policy, with approvers where the policy requires them.

In development
Planned

Agent Manager and Agent Marketplace.

As more AI agents go to work, organisations need to govern them together. Both are planned; details will be published here when they are ready to show.

Planned

Agent Manager

One place to register every AI agent, grant and revoke its permissions, monitor what it does, and switch it off.

Planned

Agent Marketplace

A catalogue of governed AI agents for regulated work: screening, identity, fraud and security, each with its guardrails and audit built in.

Principle

Governed by design

Every agent in the Manager or the Marketplace follows the same rule: it can advise and propose; approved rules and people authorise.

The line they share

What no Saolix AI agent can do.

Decide

Verdicts, holds and responses come from versioned rules or approved policy, and from people. A copilot's output is advisory.

Override a real hit

Where a copilot's suggestion conflicts with the rules, the rules win, and the case goes to a person.

Become essential

Switching a copilot off removes the help, not the correctness. Every decision still comes out the same.

How the line is enforced in Eagle →
Contact · Saolix GroupPeople. Products. Possibilities.

Your next chapter.
Let’s build it together.

Explore a product, discuss a project, or find the right expertise. Tell us where you want to go. We’ll help you take the next step.

Enterprise softwarePrivate AIEngineering & consulting