As enterprises put AI agents to work, they face a challenge: How can they give those systems more autonomy while keeping AI agent guardrails in place?