Communication to Completion: Modeling Collaborative Workflows with Intelligent Multi-Agent Communication

TL;DR

The paper addresses the challenge of coordinating teams of LLM agents on long-horizon tasks by treating communication as a controllable resource. It introduces the Alignment Factor ($AF$) and the Sequential Action Framework (SAF) to connect communication decisions directly with task progress, supported by intention-based agent reasoning and hierarchical DAG task management. Through extensive simulations on coding-like tasks across varying complexity and team sizes, the approach reduces completion time by about 40% and achieves high efficiency while maintaining manageable communication costs, with scalable performance and clear interpretable coordination patterns. This work provides a theoretical and practical framework for measuring and optimizing communication in multi-agent systems, offering actionable insights for designing scalable, adaptive collaborative AI systems.

Abstract

Teamwork in workspace for complex tasks requires diverse communication strategies, but current multi-agent LLM systems lack systematic frameworks for task oriented communication. We introduce Communication to Completion (C2C), a scalable framework that addresses this gap through two key innovations: (1) the Alignment Factor (AF), a novel metric quantifying agent task alignment that directly impacts work efficiency, and (2) a Sequential Action Framework that integrates stepwise execution with intelligent communication decisions. C2C enables agents to make cost aware communication choices, dynamically improving task understanding through targeted interactions. We evaluated C2C on realistic coding workflows across three complexity tiers and team sizes from 5 to 17 agents, comparing against no communication and fixed steps baselines. The results show that C2C reduces the task completion time by about 40% with acceptable communication costs. The framework completes all tasks successfully in standard configurations and maintains effectiveness at scale. C2C establishes both a theoretical foundation for measuring communication effectiveness in multi-agent systems and a practical framework for complex collaborative tasks.

Figures (7)

• Figure 1: Communication to Completion architecture: a five-layer stack simulation engine
• Figure 2: C2C timeline on a software bug-fix task. Step 1: the manager decomposes the task and assigns subtasks. Step 2: workers execute intentions and produce interim results. Communications update the Alignment Factor (AF); higher AF increases work efficiency and reshapes subsequent actions (green feedback). Vertical dashed dividers denote discrete timesteps in the Sequential Action Framework; the cycle repeats until completion (Step N).
• Figure 3: Analysis of the Alignment Factor (AF) dynamics. (a) Average alignment improves over successive communication rounds across all task complexities. (b) High intent communications like MEETINGS and HELP requests yield the largest gains in alignment. (c) The communication heatmap reveals a manager centric coordination pattern, with the manager (M1) acting as the central hub.
• Figure 4: Effect of task complexity under C2C (8/24/40 hours) and team configuration (1M+4W, 1M+8W, 1M+16W). (a) Completion time increases with task complexity; larger worker pools shorten time with diminishing returns. (b) Messages per hour rise with both complexity and team size. (c) Help messages dominate across settings; clarification and meeting shares grow with complexity, and progress updates appear from medium upward. (d) Communication channels shift from chat (simple) toward email (complex), with a modest increase in meetings.
• Figure 5: Prompt for the manager to decompose a give task.