AI readiness and agents
How do I stop AI from acting outside my business rules?
Put hard limits around the tools, fields, amounts, actions, and conditions the AI can use. Business rules should be enforced by the surrounding system, not trusted to the AI prompt alone. Send anything outside those limits to a person.
What to check
- Which actions are allowed?
- Which limits should be enforced outside the AI model?
- What happens when a request falls outside the rules?