VI
Talk · 20 min

Cut costs, boost efficiency with Multi AI Agents.

Learned from real-world use cases.

A breakdown of how Expand VN applies AI to control and automate their entire workflow — preventing losses and waste while boosting efficiency.

7 real use cases already deployed successfully · From theory to practice

Speaker Khoa Nguyễn
Company Snappp.ai
Duration 20 min
01 —

Theory · 5 keywords to remember

AI Agent LLM + tools + making decisions and taking action on its own. Not a chatbot.
Single Agent One agent doing one specialized job.
Multi-Agent Multiple agents working together, each with its own task.
Orchestrator The "conductor" agent — coordinates the other agents so they work in sync.
Human-in-the-Loop A human signs off at the high-risk points. AI does 90%, the human approves 10%.
02 —

Data structure inside a business

Collect → Analyze → Use
01 / TEXT
Text
Zalo messages, email, internal chat
02 / IMAGE
Images
Product photos, warehouse, orders, invoices
03 / AUDIO
Recordings
Internal meetings, customer calls, shift recordings
04 / DOCUMENT
Documents
Contracts, quotes, PDF, Word, Excel
05 / API
Platform APIs
Facebook, Shopee, TikTok, CRM, accounting
⌁ Event-based
Every event in the business = a data source. Design everything around events so no data slips through the cracks.
03 —

Use cases · 7 real cases

→
Agent that reads and analyzes Zalo
Automatically reads customer messages, classifies them into price inquiry / complaint / product interest, and alerts sales in time.
→
Agent that analyzes images in the business
Reads order photos, warehouse photos, invoice photos — extracts the numbers automatically.
→
Agent that analyzes recordings
Speech-to-text on internal meetings and customer calls, then summarizes and pulls out the key points.
→
AI Agent for price monitoring
Automatically tracks product prices on Shopee, TikTok Shop and Lazada — competitor prices, market swings, alerts and timely pricing recommendations.
→
Running the factory with AI
An agent monitors production through cameras and sensors — tracks line KPIs, detects anomalies, and sends realtime alerts to managers.
→
Knowledge Base for accounting
The agent learns all of the company's documents, invoices and tax rules → answers any accounting question, looks things up fast, and helps prepare reports.
→
Multi-Agent end-to-end
Multiple agents working together from data collection → analysis → reporting → auto-publishing posts, with an Orchestrator coordinating the whole flow.
04 —

Bonuses

✦ Orchestrator Agent pre-configured — ready to use, no code needed
✦ $5 LLM credit to try out the agents
✦ Multi-Agent template matching the exact use case demoed in the talk
All keywords
AI Agent Single Agent Multi-Agent Orchestrator Human-in-the-Loop Event-based Collect – Analyze – Use Text Image Audio Document API Knowledge Base