What are policies?

While it's appropriate to steer some behavior through natural language, enforcing business rules through natural language is dangerous. You don't want to build Agentic AI automations by hoping an AI model understand your rules.

An LLM's understanding of instructions can be misunderstood, or worse, jailbroken. Further, the more policies you add, the less likely your AI agent will reliably enforce policies (Lost in the Middle problem).

Moveworks Policy Validators enforce & guarantee that your business rules are always respected accurately — no matter how many you add.

Learn more about our approach to policy validators here.

Confirmation Policies

In Plugins under Basic Info, you can specify if User consent required before execution? to require policies

When enabled, this policy guarantees that no matter what the user says, they always get a chance to confirm the exact action that's being taken.

Other Policies

We are releasing other policy types that you can apply to your AI agents (like input validators & slot prerequisites). Follow in our community to be the first to use these new capabilities.