AI agent policy engine
Business rules should not depend on an LLM remembering them. A policy engine keeps them explicit, versioned and deterministic.
What is a policy engine for AI agents?
A policy engine for AI agents stores business rules in a structured form and evaluates them against a proposed action and its evidence. The result is deterministic: the same action with the same evidence always returns the same decision.
What a policy contains
- Action type it applies to, such as refund.create or crm.record.delete
- Required evidence before any decision can be made
- Autonomous limits the agent may act within
- Approval bands and the role that approves each band
- Hard blocks that no approval can override
- Duplicate and timing conditions
Written in business language
Policies are meant to be readable by the finance lead or the head of support who owns the rule, not only by the team that integrates the API.
A refund policy reads as a short list of statements: allow up to 100 EUR, manager approval to 500 EUR, finance approval above that, never when a chargeback exists, require invoice status paid.
Deterministic by design
Evaluation is code, not generation. When evidence is missing, the engine returns INSUFFICIENT EVIDENCE instead of assuming. That is the difference between a control and a suggestion.
Questions
Frequently asked questions
Give your agents freedom without giving them a blank cheque.
Tell us what your agents are allowed to do today. We will review whether Gatewise fits your agent workflow.
Works with your agents and existing business tools