AFK.Monks
.monks
Project Nature:
Solution
Location:
The Netherlands
Year:
2024







Key Words:
Python, Figma, Human Centric Design, XAi for Empathy, Muti-Agents System, Explainable AI (XAI), Human-AI Interaction
Let's Talk 🙃
Description:
AFK. Monks - Implementation of Large Language Model (LLM) Powered Muti-agents system
UX Design
AFK.Monks is a graduation research paper developed in collaboration with ".monks*" (formerly known as MediaMonks). This project addresses the challenges faced by project managers when they are "out of office," particularly the accumulation of unorganized communication that complicates follow-up processes. By leveraging generative AI, we aimed to reduce the cognitive load on project managers by developing a multi-agent system that automatically processes emails and enhances situational awareness.
The system was designed using a Human-Centered Design framework, involving ideation, implementation, and evaluation phases. Key functionalities were identified and refined through iterative prototyping and user feedback. The design of AFK.Monks adheres to situational awareness feedback loops, providing the right amount of information and affordance to help users navigate and interact with the AI system effectively.
Although not all design criteria were met due to limited access to testing data, the overall concept received positive feedback from testers. The results indicate a preference for functionality that improves understanding, with less emphasis placed on AI explainability. Further research and iterations are necessary to address these gaps and adapt the system for real-world office environments.
System Design
In 2023, Large Language Model (LLM) start be widely used within software system. In this master degree thesis, Joseph discussed an novel approach of using multi-agent frameworks to create automated software that integrate LLM into existing commercial infrastructure. Research suggested that the combined output of several AI agents outperforms a single agent in accuracy and reliability for complex tasks. The AFK.Monks use this approach and create specific Ai agents to finish individual tasks and the combined output will be passed to the next agents to be further processed through out the pipe line. Each Agents can use different tools and have access to different knowledge specific for their assigned tasks. Implementing such a system within existing company infrastructure can reduce friction and accelerate the adoption of LLM applications. AFK.Monks aims to offer a new perspective on the traditional “out-of-office” setting for project manager, enhancing users' situational awareness of their projects.