A Survey on Multi-Agent Workflow Automation: Methods, Taxonomies, and Future Directions

PhD Qualifying Examination


Title: "A Survey on Multi-Agent Workflow Automation: Methods, Taxonomies, and 
Future Directions"

by

Mr. Wenshuo ZHANG


Abstract:

The rapid advancement of Large Language Models (LLMs) has catalyzed the 
transition from single-agent capabilities to collaborative Multi-Agent 
Systems (MAS). However, the construction of effective agentic workflows 
currently relies heavily on manual Flow Engineering—a heuristic-driven 
process that is labor-intensive, brittle, and difficult to scale. To address 
these challenges, a new research frontier has emerged: Multi-Agent Workflow 
Automation (MAWA), which aims to autonomously discover, generate, or evolve 
optimal system architectures.

In this survey, we provide a comprehensive review of the MAWA landscape, 
analyzing 22 state-of-the-art publications from 2020 to 2025. We first 
formally define the MAWA problem as the holistic joint optimization of agent 
profiles, interaction topologies, and environmental protocols, distinguishing 
it from component-level optimizations such as isolated prompt tuning. We 
propose a novel hierarchical taxonomy that classifies existing approaches 
into four distinct categories: (1) Optimization-based Structure Discovery, 
which formulates design as a search problem within discrete spaces using 
algorithms like MCTS and evolutionary strategies; (2) Learning-based 
Structure Generation, which leverages deep learning and generative models to 
synthesize workflows directly from task descriptions; (3) Dynamic & 
Adaptive Orchestration, where workflow structures emerge or mutate at runtime 
based on feedback and complexity; and (4) System Frameworks & Robustness, 
which focuses on the stability and accessibility of automated infrastructure. 
Finally, we identify critical gaps in current research—including the 
need for interpretable generation and outline promising directions for the 
evolution of AI-designed AI systems.


Date:                   Friday, 19 December 2025

Time:                   9:00am - 11:00am

Venue:                  Room 2128A
                        Lift 19

Committee Members:      Prof. Huamin Qu (Supervisor)
                        Prof. Raymond Wong (Chairperson)
                        Dr. Arpit Narechania