DEV Community

Hieu Luong
Hieu Luong

Posted on • Originally published at himitek.com

Case Study: How a B2B Wholesale Distributor Saved 80% of Zalo Order Entry Time with AI Agent

#himitek #technology #saas

Risk Diagnosis: When Sales Admins Become Manual Data Entry Machines

In Vietnam, up to 90% of wholesale transactions between distributors and dealers (FMCG, construction materials, spare parts...) still happen via Zalo. Instead of using structured ordering portals, dealers prefer sending casual messages like: "Send me 5 boxes of blue detergent, 2 red packs, urgent delivery please" or sending a photo of a handwritten note. Sales Admins must manually read, look up SKU codes in Excel files, and type them into accounting/ERP software. This is a dangerous operational bottleneck.

Operational Impact: One Typo, Heavy Logistics Costs

This manual order processing leads to 3 immediate financial leaks for business owners:

  • Wasted Payroll: Qualified staff who should be doing customer care and up-selling are turned into data entry typists 8 hours a day. During peak seasons, overload forces them into endless overtime.
  • Return Logistics Costs: Just one mistyped SKU character (e.g., 500ml bottle instead of 1L) causes incorrect deliveries. The distributor bears all two-way shipping and stock handling costs.
  • Customer Churn: Slow and incorrect deliveries damage trust. Dealers will quickly switch to competitors who respond and deliver faster.

3-Step Solution: Automating Order Entry with AI Agent

HimiTek deploys an AI Sales Admin Agent that works behind the scenes to normalize Zalo data and push it directly to your ERP in 3 steps:

Step 1: Capture Zalo Messages. Webhooks collect text messages or images from Zalo OA.

Step 2: Extract & Normalize SKUs. The AI processes natural language, mapping slang terms ("blue detergent", "red pack") to the exact standard SKUs in your system.

Step 3: Create Draft Orders. The AI pushes normalized data into the ERP via API. Sales Admins only need to review and click approve.

Here is a Python code snippet demonstrating how the AI Agent processes unstructured Zalo messages into clean SKU data:

import openai
import json

def normalize_order(zalo_text):
# Actual SKU catalog in warehouse
sku_catalog = {
\'bot giat xanh\': \'BG-XANH-01\',
\'loc do\': \'BG-DO-02\'
}

prompt = f\'\'\'

Analyze the following order message and return JSON format.

Message: \'{zalo_text}\'

Map with SKU catalog: {sku_catalog}

Return result as JSON list: [{{\'sku\': \'...\', \'quantity\': ...}}]

\'\'\'


response = openai.ChatCompletion.create(

    model=\'gpt-4\',

    messages=[{\'role\': \'user\', \'content\': prompt}]

)

return json.loads(response.choices[0].message.content)
Enter fullscreen mode Exit fullscreen mode




Real test

raw_msg = \'Cho anh 5 thung bot giat xanh, 2 loc do\'
print(normalize_order(raw_msg))

Output: [{\'sku\': \'BG-XANH-01\', \'quantity\': 5}, {\'sku\': \'BG-DO-02\', \'quantity\': 2}]## Optimize Your Revenue with HimiTek

Stop letting your Sales Admins burn out on repetitive tasks. The AI Agent solution cuts order entry time from 15 minutes to 30 seconds and eliminates 99% of typing errors. Contact HimiTek today to build a custom AI assistant tailored to your distribution workflow!

Top comments (0)