Building a
Multi-Agent Office Assistant with RAG & Agents
Dr. Tamer Arafa,
School of Information Technology and Computer Science, Nile
University, Egypt
Dr. Ghada Khoriba,
School of Information Technology and Computer Science, Nile
University, Egypt
Mr. Ahmed Tamer,
Graduate School of Information Science, University of Hyogo,
Japan
The emergence
of AI multi-agent systems marks a significant step forward in
the evolution of intelligent automation, as they enable multiple
specialized agents to collaborate and coordinate in solving
complex, dynamic problems that a single model alone may not
handle effectively. By distributing tasks across agents with
distinct roles, such as retrieval, reasoning, summarization, or
task execution, multi-agent systems mirror human teamwork, where
cooperation and delegation drive efficiency and accuracy. This
collaborative structure enhances scalability, adaptability, and
robustness, making it possible to tackle multi-faceted
real-world challenges, from managing office workflows to
supporting decision-making in high-stakes domains like
healthcare and finance. Ultimately, AI multi-agent systems are
useful because they combine the complementary strengths of
different models and processes, yielding more reliable,
context-aware, and efficient outcomes than traditional
single-agent solutions.
The workshop
“Building a multi-agent office assistant with RAG &
agents” offers a unique opportunity for participants
to gain hands-on experience with some of the most cutting-edge
approaches in applied artificial intelligence.
Retrieval-Augmented Generation (RAG) has emerged as a powerful
method for improving the accuracy, reliability, and contextual
grounding of large language model outputs by integrating them
with domain-specific knowledge bases. Building upon this
foundation, the workshop extends into the realm of multi-agent
systems, where multiple specialized AI agents collaborate to
perform complex tasks such as answering context-sensitive
questions, summarizing large volumes of documents, and
automating follow-up actions within an office environment.
Participants
will not only learn the theoretical underpinnings of RAG and
multi-agent orchestration but will also engage in practical
exercises where they design, implement, and test a fully
functional office assistant. By the end of the day, attendees
will have developed a system capable of demonstrating the
synergy between retrieval pipelines and collaborative agents.
Observed skills are directly transferable to real-world
applications in enterprise productivity, knowledge management,
and intelligent workflow automation.
The workshop will take
place on Friday 16 January
2026 and the tentative
schedule is as follows:
|
Session |
Key
Activities |
Dataset
Used / Outcome |
|
|
09:00 – 09:30 |
Setup & Intro |
Environment
setup, overview of LLMs, RAG, and multi-agent roles. |
— / Ready to
start |
|
09:30 – 10:30 |
Concepts |
Deep dive into
RAG pipelines and agent orchestration. Use cases for
office automation. |
— / Core
understanding |
|
Coffee Break |
|||
|
10:45 – 12:30 |
Hands-On RAG |
Load Office Mini
Dataset into FAISS/Chroma. Build a retrieval Q&A
pipeline with LangChain. |
HR policies,
emails, reports / Working RAG |
|
Lunch |
|||
|
13:30 – 15:00 |
Advanced RAG |
Add
summarization, contextual retrieval, and multi-document
Q&A. Improve retrieval quality. |
Emails + meeting
transcripts / Smarter RAG system |
|
Coffee Break |
|||
|
15:15 – 16:45 |
Multi-Agent
Assistant |
Orchestrate
Retriever, Summarizer, Planner with LangGraph/AutoGen.
Demo: summarize meeting → draft follow-up email. |
Combined
datasets / Prototype office assistant |
|
16:45 – 17:30 |
Wrap-Up & Q&A |
Scaling,
deployment, security, feedback, networking. |
— / Next steps |
