What Are Agents?
Agents are the core of Velatir’s compliance engine. Each agent specializes in a distinct area of risk, evaluating every trace that flows through your organisation. Together, they cover data protection, access control, HR compliance, brand safety, and commercial risk. You don’t need to trigger agents manually. They evaluate every trace in real time, applying the rules you’ve configured to determine whether a trace should be allowed, escalated, or blocked.How Agents Work
Every time someone in your organisation interacts with an AI service, that interaction is captured as a trace. Each agent independently evaluates the trace against its own set of rules and produces an intent: allow, block, or escalate. The agent’s role then determines what action is actually taken. This means you get layered protection. A single trace might be flagged by Data Protection for containing personal information and by Commercial Advisor for referencing pricing terms. Each agent contributes its own perspective.Available Agents
Data Protection
Detects personal data exposure, financial information leaks, credential sharing, and intellectual property risks. Central to GDPR and privacy compliance.
Gatekeeper
Controls which AI services your organisation can use and under what conditions. Enforces usage rules through allowlists and blocklists.
People Operations
Evaluates AI usage in HR processes like hiring, performance reviews, and compensation decisions. Surfaces risks related to bias and employment law.
Brand Guardian
Monitors brand safety and reputation risk. Detects brand inconsistency, competitor information sharing, and messaging that could affect your public image.
Commercial Advisor
Evaluates business and financial risk. Catches unauthorised commitments, pricing exposure, and commercially sensitive information before they leave your organisation.
Roles and Intents
Every agent operates with a role that determines how much authority it has. It also produces an intent that expresses what it thinks should happen with a given trace. Roles control the agent’s level of authority:- Observer watches and logs findings without taking action.
- Enforcer can block traces and require human approval before proceeding.
- Allow means the trace is compliant. It proceeds normally.
- Block means the trace poses a risk and should be stopped.
- Escalate means the trace needs human review.
Roles and Intents
How roles and intents combine to determine agent actions.
Configuring Agents
How to set up agents at the organisation and workspace level.