BeautyGuard: Designing a Multi-Agent Roundtable System for Proactive Beauty Tech Compliance through Stakeholder Collaboration

TL;DR

BeautyGuard addresses the challenge of enterprise compliance in beauty tech by introducing a multi-agent LLM roundtable that mirrors organizational roles to surface early, explainable, and actionable guidance. The four role-specific agents—Legal Interpreter, Rule Checker, Precedent Researcher, and Risk Planner—work in streaming, progressively revealing their reasoning to support proactive mitigation rather than binary gating. A formative study with six experts informs design (DC1–DC5), and a user study (N=6) demonstrates high usability (SUS 77.5/100) and low workload across most NASA-TLX dimensions, with results showing effective knowledge preservation, preference for information augmentation, and good alignment with existing organizational structures. The work provides design principles for human–AI collaboration in regulated domains and suggests generalizability to other industries and governance processes beyond beauty tech.

Abstract

As generative AI enters enterprise workflows, ensuring compliance with legal, ethical, and reputational standards becomes a pressing challenge. In beauty tech, where biometric and personal data are central, traditional reviews are often manual, fragmented, and reactive. To examine these challenges, we conducted a formative study with six experts (four IT managers, two legal managers) at a multinational beauty company. The study revealed pain points in rule checking, precedent use, and the lack of proactive guidance. Motivated by these findings, we designed a multi-agent "roundtable" system powered by a large language model. The system assigns role-specialized agents for legal interpretation, checklist review, precedent search, and risk mitigation, synthesizing their perspectives into structured compliance advice. We evaluated the prototype with the same experts using System Usability Scale(SUS), The Official NASA Task Load Index(NASA-TLX), and interviews. Results show exceptional usability (SUS: 77.5/100) and minimal cognitive workload, with three key findings: (1) multi-agent systems can preserve tacit knowledge into standardized workflows, (2) information augmentation achieves higher acceptance than decision automation, and (3) successful enterprise AI should mirror organizational structures. This work contributes design principles for human-AI collaboration in compliance review, with broader implications for regulated industries beyond beauty tech.

Figures (7)

• Figure 1: Workflow of the AI Roundtable System: The process begins with a user uploading a proposal (1), which is then parsed by an LLM (2). It proceeds to the core AI Roundtable Review, where four specialized agents independently analyze the proposal, with their outputs visible to the user (3). A consolidated report is generated (4), after which the user can ask questions (5) to trigger a new round of agent discussion for iterative refinement (6).
• Figure 2: User Journey of the AI Roundtable System: A Four-Stage Multi-Agent Review Process.A walkthrough of a sample query, from initial proposal submission, through the multi-agent analysis, to the final consolidated report that provides a clear risk summary and actionable recommendations for improvement.
• Figure 3: Progressive Disclosure Flow: Illustrating the progressive disclosure of UI interface from case submission (a) through individual expert analyses (b-d) to the final consolidated report (e).
• Figure 4: Roundtable Interface Design. After a proposal is submitted (a), the user is presented with an overview (b, c, d) of expert analyses. The design allows users to explore the detailed report from each agent (Legal, Case Research, Checklist Review) independently, facilitating a transparent and granular review process.
• Figure 5: User Study Procedure.illustrating the process from initial interviews and system demonstration to the collection of mixed-methods data on usability and user experience.