Orchestration-Free Customer Service Automation: A Privacy-Preserving and Flowchart-Guided Framework

Mengze Hong; Chen Jason Zhang; Zichang Guo; Hanlin Gu; Di Jiang; Li Qing

cs.CLFeb 17, 2026

Orchestration-Free Customer Service Automation: A Privacy-Preserving and Flowchart-Guided Framework

Authors: Mengze Hong, Chen Jason Zhang, Zichang Guo, Hanlin Gu, Di Jiang, Li Qing

Organizations: Hong Kong Polytechnic University, Hong Kong, China · AI Group, WeBank, Shenzhen, China

Abstract

Customer service automation has seen growing demand within digital transformation. Existing approaches either rely on modular system designs with extensive agent orchestration or employ over-simplified instruction schemas, providing limited guidance and poor generalizability. This paper introduces an orchestration-free framework using Task-Oriented Flowcharts (TOFs) to enable end-to-end automation without manual intervention. We first define the components and evaluation metrics for TOFs, then formalize a cost-efficient flowchart construction algorithm to abstract procedural knowledge from service dialogues. We emphasize local deployment of small language models and propose decentralized distillation with flowcharts to mitigate data scarcity and privacy issues in model training. Extensive experiments validate the effectiveness in various service tasks, with superior quantitative and application performance compared to strong baselines and market products. By releasing a web-based system demonstration with case studies, we aim to promote streamlined creation of future service automation.

Figures & tables

Figure 1. Comparison of agent orchestration-based (top) and flowchart-guided (bottom) system for service automation.

Figure 2. System overview: orchestration-free and privacy-preserving customer service automation driven by 1) flowchart-augmented task coordination and 2) flowchart-guided synthetic data generation for model distillation.

Figure 3. Decentralized flowchart distillation framework.

Method	Attraction		Hotel		Restaurant		Taxi		Train
Method	CPC	UMR	CPC	UMR	CPC	UMR	CPC	UMR	CPC	UMR
Human-Annotated	0.86	0.8176	0.76	0.7966	0.80	0.8071	0.92	0.6728	0.84	0.7183
Abstraction-Based	0.80	0.7958	0.64	0.8103	0.74	0.8044	0.90	0.6843	0.74	0.7987
Iterative Construction	0.92	0.9436	0.80	0.8209	0.90	0.9087	0.88	0.8221	0.88	0.8277

Table 1. Coverage evaluation of flowchart construction methods on the MultiWoZ 2.0 dataset.

Unweighted DIC
	MultiWOZ		SGD
Method	# Dialog	# Utterances	# Dialog	# Utterances
Greedy	57	1012	23	318
LP-Rounding	67	1140	37	596
ILP	43	766	21	344
Weighted DIC

Table 2. Comparison of dialogue selection optimizations. The best performing method (i.e., lowest cost) is marked in blue .

Training-based Approaches
	MultiWOZ			SGD
Method	Inform	Success	Book	Inform	Success
SimpleTOD	84.4	70.1	-	12.7	9.8
UBAR	83.4	70.3	-	-	-
GALAXY	85.4	75.7	-	-	-
Mars	88.9	78.0	-	-	-

Table 3. Goal completion evaluation on MultiWoz 2.0 and SGD. The best performing method is marked in blue .

Figure 4. Comparison of Exemplary Rate and Completion Rate for four customer service outbound automations.

Figure 5. Human evaluation of proactive outbound systems across five technical aspects (0 = worst, 4 = best).

Figure 6. Comparison of sensitive data memorization measured by user preference leakage (lower is better).

Tasks	Description
Verification	Validates correct, incorrect, or missing inputs with robust error handling.
Credit Limit	Guides users through credit limit increases via online banking or mobile apps.
Account Inquiry	Manages balance inquiries and unknown transaction investigations with accurate data retrieval and issue resolution.
Travel Notification	Facilitates the setup of overseas travel alerts with app guidance or direct assistance.
Open Questions	Addresses queries on wallet deposits, age limits, loan rates, and product recommendations.

Table 4. Case scenarios for system demonstration.

Figure 7. System demonstration: flowchart-guided customer service automation for banking applications.

Appendix figures & tables1 asset

Supplementary material from the paper’s appendix.

Appendix

Figure 8. Example of task-oriented flowchart.

Customer Support Automation

Customer support automation uses AI agents, retrieval, and workflow tools to resolve inquiries, assist service staff, and streamline customer operations.

See all 35 papers

Autonoma: A Hierarchical Multi-Agent Framework for End-to-End Workflow Automation
Feb 27, 2026Eslam Reda, Maged Yasser, Sara El-MetwallyWorkflow Automation Multi-Agent Orchestration
Artificial Intelligence Program, Faculty of Computers and Information, Mansoura University, Mansoura 35516, Egypt. · Computer Science Department, Faculty of Computers and Information, Mansoura University, Mansoura 35516, Egypt.
ChromaFlow: A Negative Ablation Study of Orchestration Overhead in Tool-Augmented Agent Evaluation
May 13, 2026Tarun MittalAI Agent Reliability LLM Agent Orchestration
Octave-X
In-Context Prompting Obsoletes Agent Orchestration for Procedural Tasks
Apr 30, 2026Simon Dennis, Michael Diamond, Rivaan Patil +2LLM Prompting LLM Agent Orchestration
i14, University of Melbourne · i14

Feb 27, 2026J/K move · Enter open · S save

Autonoma: A Hierarchical Multi-Agent Framework for End-to-End Workflow Automation

Eslam Reda, Maged Yasser, Sara El-Metwally

Artificial Intelligence Program, Faculty of Computers and Information, Mansoura University, Mansoura 35516, Egypt. · Computer Science Department, Faculty of Computers and Information, Mansoura University, Mansoura 35516, Egypt.

Workflow Automation Multi-Agent Orchestration Multi-Agent Systems

Open paper AskPDF arXiv

The increasing complexity of user demands necessitates automation frameworks that can reliably translate open-ended instructions into robust, multi-step workflows. Current monolithic agent architectures often struggle with the challenges of scalability, error propagation, and maintaining focus across diverse tasks. This paper introduces Autonoma, a structured, hierarchical multi-agent framework designed for end-to-end workflow automation from natural language prompts. Autonoma employs a principled, multi-tiered architecture where a high-level Coordinator validates user intent, a Planner generates structured workflows, and a Supervisor dynamically manages the execution by orchestrating a suite of modular, specialized agents (e.g., for web browsing, coding, file management). This clear separation between orchestration logic and specialized execution ensures robustness through active monitoring and error handling, while enabling extensibility by allowing new capabilities to be integrated as plug-and-play agents without modifying the core engine. Implemented as a fully functional system operating within a secure LAN environment, Autonoma addresses critical data privacy and reliability concerns. The system is further engineered for inclusivity, accepting multi-modal input (text, voice, image, files) and supporting both English and Arabic. Autonoma achieved a 97% task completion rate and a 98% successful agent handoff rate, confirming its operational reliability and efficient collaboration.

Abstract

Customer Support Automation

Explore similar work

Autonoma: A Hierarchical Multi-Agent Framework for End-to-End Workflow Automation

ChromaFlow: A Negative Ablation Study of Orchestration Overhead in Tool-Augmented Agent Evaluation

In-Context Prompting Obsoletes Agent Orchestration for Procedural Tasks