Organizational Control Layer: Governance Infrastructure at the Execution Boundary of LLM Agent Systems
2026-06-03 • Multiagent Systems
Multiagent Systems
AI summaryⓘ
The authors study how large language model (LLM) agents can cause problems when their generated suggestions directly control actions in the real world. They propose separating the step where an agent suggests actions from the step where those actions are actually done, using a system called the Organizational Control Layer (OCL). OCL checks and controls actions before execution to keep things safe without changing the language model itself. Their tests show OCL greatly reduces unsafe actions and improves successful outcomes, though stricter controls can sometimes limit flexibility. This means LLM agents need strong checks between their ideas and their real actions to work safely in complex settings.
large language modelsagent systemsaction governancepolicy enforcementexecution boundarymulti-agent interactionsafety-utility tradeoffOrganizational Control Layeradversarial negotiationdeployment infrastructure
Authors
Tianyu Shi, Yang Mo, Yiou Liu, Zhuonan Hao, Yin Wang, Wenzhuo Hu, Nan Yu, Meng Zhou, Jiangbo Yu
Abstract
LLM-based agents are increasingly deployed in workflows where generated outputs may directly trigger state-changing actions. This creates an execution-boundary problem: proposed actions must be governed before they are executed. We study this problem through economically consequential multi-agent interactions and argue that deployment-grade agent systems should separate proposal generation from environment-facing execution. To operationalize this principle, we introduce the Organizational Control Layer (OCL), a model-agnostic governance infrastructure that intercepts generated actions before execution through policy enforcement and escalation, without modifying the underlying LLM generator. We evaluate OCL on adversarial buyer--seller negotiation environments adapted from AgenticPay. Across multiple frontier LLM backends, OCL reduces unsafe executions from 88% to near-zero while increasing valid success from 12% to 96%. Results further reveal a safety--utility tradeoff: strict governance improves compliance and reliability against policy and constraint violations, but can reduce flexibility in tightly constrained markets. These findings suggest that deployment-grade LLM agent systems require explicit governance at the boundary between language generation and executable actions. The source code is available at: https://github.com/SHITIANYU-hue/amai_ocl