Organizational Control Layer: Governance Infrastructure for Mixed Human-AI Economic Systems

TIANYU SHI, Yang Mo, Zhuonan Hao, Zhoumeng, Wenzhuo Hu, Nan Yu, Fucheng Deng, Jiangbo Yu

Keywords: LLM agents; agent evaluation; multi-agent systems; AI governance; economic negotiation

TL;DR: We propose an Organizational Control Layer that governs LLM agents in economic workflows by checking, revising, blocking, or escalating proposed actions before execution.

Abstract:

LLM-based agents are increasingly used in economic interactions such as negotiation, procurement, and conversational commerce, where generated responses may correspond to prices, terms, or commitments. This raises a governance problem: candidate actions should be checked against constraints and policies before they are executed. We present the Organizational Control Layer (OCL), a model-agnostic layer that separates action generation from action execution by routing candidate actions through approval, revision, blocking, or escalation. We evaluate OCL in a modified AgenticPay-style negotiation setting with adversarial buyer scenarios. We find that unguided agents can achieve high agreement rates while still producing unsafe or invalid outcomes. OCL improves compliant negotiation outcomes by checking actions before execution and
recording the resulting control decisions. These results support a simple design principle: economic LLM agents should be evaluated and governed at the point of execution, not only at the point of generation.

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Submission Number: 14