Beyond intelligence: Why trust will define enterprise AI

When an AI agent pays an invoice it shouldn't have, the liability is still yours. The auditor still asks why and the board still expects someone to answer for it. That is why 61% of organizations only let AI execute business decisions with a human in the loop, and only 26% say their governance is ready to change that.

In this IDC Spotlight, Kevin Permenter, Senior Research Director for Financial Applications and Agents at IDC, explains what has to be true before finance can let AI act on its own, and where in your architecture those controls belong.

What's inside

• What IDC requires before AI can act on its own: one policy enforced the same way in every system, a decision record an auditor can follow, and limits set in advance for what AI decides versus what escalates
• A worked example of an invoice 8% over its PO, and why procurement, AP, and the ERP each give the agent a different answer
• Why IDC argues governance belongs at the orchestration layer instead of inside each tool or agent
• A four-stage path for widening AI's authority as the decision record proves it follows policy
• IDC's advice for evaluating AI platforms on policy adherence and auditability alongside automation rates