300 vendor messages a day. Today's prices by 7:30am.
This supplier serves restaurants and hotels in Vietnam and Cambodia and buys from 100+ vendors. Four people typed vendor prices into Excel every morning. Now MIDAS reads the vendor messages, checks each order, drafts purchase orders for missing stock and gives each person their next task. The team checks and approves.
Client
Food supplier to restaurants & hotels
Where
Vietnam & Cambodia
Scale
100+ vendors · ~300 messages a day
Setup
An engineer on site for 1 day
How did a food supplier stop typing vendor prices by hand?
This supplier serves restaurants and hotels in Vietnam and Cambodia and buys from 100+ vendors. Four people typed vendor prices into Excel every morning. Now MIDAS reads the vendor messages, checks each order, drafts purchase orders for missing stock and gives each person their next task. The team checks and approves.
Query fit: food supplier; vendor price list; supplier quotes from email and Telegram; purchase orders for missing stock; order fulfillment tasks.
Price list
Live
was ready at noon
To enter and check an order
~1 min
from ~15 min
From stock shortage to purchase order
Minutes
was the next day
Morning data entry
~30 min
from 4 people × ~4 hrs
By noon, the price list was already old.
Vendor prices and stock arrived by email and Telegram. Restaurant orders came through messages too. Staff copied the numbers into Excel by hand.
Four people typed all morning.
About 300 vendor messages arrived each day. The price list was ready at noon, when some prices were already old.
Nobody could trace a wrong price.
Vendors changed prices after 8am. Typing mistakes went unnoticed. Nobody could tell who had changed a number.
Orders ran on Excel too.
Checking stock, prices and customer debt for each order took the rest of the day.
Missing stock was found too late.
Staff often found shortages while packing an order. The purchase order went out the next day. Restaurants had to wait.
Set up in 1 day.
A trader has no factory to map. An NBR engineer loaded the company's product list, vendors and customers, connected email and Telegram, and the team used MIDAS the same day.
Connect
Morning · email and Telegram groups
Load data
Morning · products, vendors, customers
Set who does what
Afternoon · buying, receiving, delivery
Train the team
Afternoon · web, phone and MIDAS chat
Vendor prices update as messages arrive.
Email and Telegram are connected to MIDAS. It reads photos, PDFs and Excel files in every vendor message. It converts box prices to prices per kg and records vendor stock. Changes over 5% need a person to check. Each price links to its original message.
Buying prices per kg, with vendor stock and the original messages. Example data.
Enter and check each order in about 1 minute.
A sales rep forwards the restaurant's message to MIDAS. It checks whether prices are below the customer's last order or debt is over the limit. It also checks for duplicate orders and unclear product names. Problems go to a manager. Orders that pass are saved right away.
A restaurant order with the checks complete. The salmon purchase is waiting for approval. Example data.
A purchase order in minutes. A task for each person.
MIDAS reserves available stock. It drafts a purchase order for missing items using this morning's best vendor price. A manager approves with one tap. MIDAS plans the work and assigns each person a task. Each task opens after the previous step is done. The order is tracked until shipped.
Order #2207: buying and delivery tasks, with the person responsible for each step. Example data.
Results after 2 weeks with MIDAS.
Price list ready
Before: Noon
After: Live
The full list is ready by 7:30am and keeps updating.
Morning data entry
Before: 4 people, ~4 hrs
After: ~30 min
One person checks the entries MIDAS marks for review.
Price typing errors
Before: ~1 in 20 lines
After: <1 in 200
No manual copying of vendor prices.
Enter and check an order
Before: ~15 min
After: ~1 min
Order checks happen before saving.
From stock shortage to purchase order
Before: Next day
After: Minutes
No wait until the next day to prepare the purchase order.
Why the team kept using MIDAS
Nobody had to learn a big ERP system.
Fix the most time-consuming job first.
Nothing changed for vendors.
They kept using email and Telegram.
People still decide.
MIDAS helps the team check. Managers decide what to approve.
Their data stays theirs and is never used to train AI.
Next: the Cambodia warehouse will use the same price list and order plans.
See MIDAS with your own orders.
Frequently asked questions.
How long did the setup take?
1 day. A trader has no factory to set up, so the main job is loading the product list. An NBR engineer did that on site and helped staff use MIDAS on real orders the same day.
What results did the food supplier see?
Price list: noon → live. Morning data entry: 4 people × ~4 hours → ~30 minutes of checking. Price typing errors: ~1 in 20 lines → under 1 in 200. Order entry: ~15 min → ~1 min. Missing stock to purchase order: next day → minutes.
What happens if MIDAS cannot read part of a message?
It marks those rows for a person to check. Rows it can read are still saved.
How does MIDAS choose the vendor for a purchase order?
MIDAS compares vendor prices. It keeps the usual vendor unless stock is unavailable or a quote from the last 30 days saves more than an extra order costs. Quantities follow the vendor's box size and minimum order. Each item shows the reason for the choice. A manager approves.
Does MIDAS handle expiry dates?
Yes. Staff record expiry dates when goods arrive. The batch that expires first ships first. Staff can trace each batch to the restaurants that received it.
Is the company's data used to train AI?
No. MIDAS uses Microsoft Foundry. Microsoft's terms and our AI provider contracts prohibit training on company data. Each person can only access what their job allows. MIDAS records who changed what.
Explore
Website: https://nbrintelligence.com