What question did this study set out to answer?

The central aim is to enhance multi-agent interaction by enabling entire agent organizations to respond to a single directive.

March 13, 2026Open Access

From One Directive to Full-Organization Action: Organizational Mirroring for Multi-Agent LLM Systems

Key Points

The central aim is to enhance multi-agent interaction by enabling entire agent organizations to respond to a single directive.
Developed an approach called organizational mirroring for LLM systems.
Implemented eight guiding principles in a production system of 18 agents across four departments.
Conducted controlled experiments over a 30-task suite with 90 runs across three communication topologies.
Achieved 3.1–5.9x greater output volume.
Received quality scores averaging 18.3 out of 20.
Identified hierarchical coordination as a major quality driver (Cohen's d=1.409).
Noted an 85.7% reduction in necessary communication links.
Found no cross-domain lexical intrusion present.
Achieved 87.5% agreement between independent evaluation layers.

Abstract

Current multi-agent frameworks adopt flat communication topologies, requiring designers to manually orchestrate each agent interaction. We present organizational mirroring, an approach that allows a single natural-language directive to mobilize an entire LLM agent organization—mirroring how a CEO's briefing cascades through a real company. The approach rests on eight principles: hierarchical delegation, independent memory, layered compression, meta-department, skill-based composition, self-evolution, replaceable executors, and real-world workflow mapping. We implement these in a production system of 18 agents across four departments on the OpenClaw framework. Controlled experiments on a 30-task suite (90 runs across three topologies) show: (1) 3.1–5.9x greater output volume; (2) quality scores of 18.3/20; (3) hierarchical coordination as primary quality driver (Cohen's d=1.409); (4) 85.7% communication link reduction; (5) zero cross-domain lexical intrusion; (6) 87.5% inter-rater agreement between independent evaluation layers. This archive contains both the English version (paper-EN.pdf) and the Chinese version (paper-CN.pdf) of the paper.

From One Directive to Full-Organization Action: Organizational Mirroring for Multi-Agent LLM Systems

Key Points

Abstract

Cite This Study