Process orders from email automatically

The order is already in the email.
So why is someone retyping it?

For wholesalers, manufacturers and suppliers whose customers order by email, with a PDF attached or just a list in the message. I build AI order processing that reads the email, extracts the lines, matches the customer and article numbers to your own system and prepares a draft order. When it is unsure, a human approves. Nothing is booked blindly.

Inboxpurchasing@baker-installations.com

Order week 43

Hi, for Thursday please: 20 carriage bolts M8 x 60, 5 boxes of black cable ties 200 mm and 10 of those white power strips from last time. Cheers, Mark

Draft order

Nothing processed yet. Nothing gets booked until someone approves.

Sample email · no language model runs here

Built for
AvitlyVTS AudiovisualDoubleStreetPub CreationsCueva ControlKattenkwaad voor ScholenErgoZorgVaatwasser LeasenVeld VisualsDe Blije BabsDe MattiesSomeWineAvitlyVTS AudiovisualDoubleStreetPub CreationsCueva ControlKattenkwaad voor ScholenErgoZorgVaatwasser LeasenVeld VisualsDe Blije BabsDe MattiesSomeWine
Sound familiar

Where order entry gets stuck.

01

An inbox full of orders every morning

Customers email their order, someone opens the attachment and retypes it line by line into the ERP. When it is busy that falls behind and the rest of the work waits.

02

Every customer a different format

One sends a PDF from their purchasing system, another an Excel file, a third types three lines in the email. A fixed import rule works for nobody.

03

Article numbers that do not match

The customer uses their own numbers or an old description. Someone has to look up which item is meant, and often only one colleague knows.

04

Typos that only show up on delivery

A ten becomes a hundred, a size gets swapped. The mistake only comes back when the customer calls to say something else was delivered.

05

Price and stock not checked

The order is in, but whether the agreed price is right and whether the item is in stock is only seen later. Then the customer has to be called after all.

06

Dependent on one person

When the colleague who enters the orders is ill or on holiday, the inbox piles up. Nobody else knows all the customers and how they order.

How it works

From email to order, with a human where it counts.

01

Read and extract

The email and the attachment are read, including a PDF or an Excel file. The model extracts the customer, delivery address, order lines, quantities and requested delivery date.

02

Match to your own data

Customer and articles are looked up in your own system, through tools on your own database or your ERP. The model does not guess an article number: what cannot be found is flagged.

03

Check

Prices come from your price list or the agreements with that customer, not from the email. Stock is checked. A deviation shows up on screen instead of in the order.

04

Prepare or ask for approval

An order that checks out on every point is ready as a draft. If the system is unsure about anything, the order goes to a review screen where a colleague reads it and approves it.

In detail
01

From email or PDF attachment to an order in your system

Getting orders from email into your ERP happens in fixed steps. For each step it is decided up front what the system may do by itself and when a human has to look. The same goes for reading an order confirmation automatically that a supplier sends you: the same steps, the other way round.

StepWhat happensWhen a human looks
1. Read the emailThe email and attachments are fetched from a dedicated inboxWhen the email does not look like an order
2. Extract linesCustomer, address, articles, quantities and delivery date from text, PDF or ExcelWhen a line is unreadable or incomplete
3. MatchCustomer and article numbers are looked up in your own systemWhen a customer or article cannot be found unambiguously
4. CheckPrice from your price list, stock from your systemWhen price or stock deviates
5. Create the orderA draft order in your ERP or own systemWhen anything was in doubt earlier, before booking
02

What happens when the system is unsure

This is the most important part, and the first thing I design. An order that lands in your ERP on a guess is worse than an order entered by hand an hour later.

So every order gets a status. If everything checks out, it is ready as a draft. If an article was not found, a price deviates or there is not enough stock, the order goes to a review screen with the original email beside it and the doubt highlighted. A colleague adjusts, approves, and only then is it booked.

  • The model never guesses an article number or customer
  • Prices come from your own system, not from the email
  • Every doubt shows up on screen with the reason
  • The original email sits beside the order in the review screen
  • What is not approved is not booked
  • A log of what happened to every email
03

Which systems it connects to

In the end the order has to land in the system you already work with. That can be Exact Online, AFAS, another package with an API or your own ERP. How that works with Exact is on Exact Online integration; for other packages see integrations.

I have built an app for a wholesaler and an ERP for a hospitality business: systems where orders come in. So I know how orders, articles and stock flow through a company like that, and where the exceptions are.

Exact Online

Customers, articles and prices are read from Exact; the approved order goes back as a sales order.

AFAS

Through the integrations AFAS offers for this. Which data is available, we look at in the first call.

Own ERP or custom system

Your own system with an API or a database I can reach. If the package has no way in at all, I tell you up front.

04

Or let customers order themselves

AI order processing is the answer when your customers keep emailing, and they often do. If you want them to order in your system themselves, with their own prices and article numbers, a B2B portal or a B2B webshop is the structural solution. It can also be both: the portal for those who want it, the AI for those who keep emailing.

If you want more than orders, such as quotes from requests or an inbox that gets sorted, that is on AI agent development.

05

What it costs and how long it takes

The build is a fixed amount you hear up front, excluding VAT. You pay for the AI model usage and hosting yourself; how much that is depends on how many orders come in per month. How pricing works is on pricing.

Order processing in an existing application · € 2,950 to € 5,500

Your orders, customers and articles are already in one system I can extend. Lead time three to six weeks.

Order processing on your ERP or Exact · € 4,500 to € 8,500

First the integration with your ERP, then the order processing on top. You hear the lead time up front.

Reviews

Clients rate me 5 stars on average.

“I am extremely happy with the result of my new website for DoubleStreet Australian Cobberdogs, made by Webframer. From first contact, Webframer radiated expertise and creativity.”
Ben VorsteveldDoubleStreet Australian Cobberdogs
“Webframer helped me enormously with the Veldvisuals website. It is not only good to spar with Bas about building the site, but also in the aftercare and whenever changes have to be made. Recommended!”
Nick in 't VeldVeldvisuals
“Great to work with. Listens carefully to what we want. When we ask for changes, they are made expertly.”
Marieke van 't WoutVan Der Linde Mode & Lingerie
“Really pleasant to work with and short lines of communication!”
Hugo BrandwagtGoogle review
“The website looks great!! Definitely recommended!”
Jayden BresserGoogle review
“Excellent! Fast service and good contact!”
Devin de WaardGoogle review

Translated from the Dutch originals on Google.

Bas Voets, founder of Webframer
Who builds it

I am Bas Voets. Since 2021 I have been building websites, client portals and business software for small and mid-sized companies, from Berkel en Rodenrijs in the Netherlands. From first call to launch you talk to me, not to an account manager.

More about meCoC 70168741 · since 2021 · Berkel en Rodenrijs
Frequently asked questions
How does AI order processing work?
Orders arrive at a dedicated email address. The system reads the email and the attachment, extracts the order lines, looks up customer and articles in your own system, checks price and stock and prepares a draft order. When in doubt, the order goes to a review screen.
What if the AI makes a mistake?
That is what the review screen is for. The system does not guess: if an article cannot be found unambiguously, a price deviates or a line is unreadable, a colleague approves the order before it is booked. Prices always come from your own system, not from what the model reads.
Is my data safe, and where is it stored?
Your application and database run inside the EU, and you get a data processing agreement. For extraction, only the email and its attachment go to the AI provider you choose. If that processing has to stay inside the EU as well, we pick a provider or region where it can.
What does it cost per month on top of the build?
You pay for AI model usage and hosting. How much that is depends on the number of orders and how long the attachments are, so I will not quote a figure that does not fit your case. In the first call I work it out from your own volume, and usage is capped with a daily limit.
Does it work with Exact Online or AFAS?
Yes, through the integrations those packages offer. Your own ERP works too, if it has an API or a database I can reach. If your package has no way in at all, you hear that in the first call.
What if every customer sends a different format?
That is exactly what AI is good at. A fixed import rule breaks with every new format; a language model reads a PDF, an Excel file and a typed list the same way. What it is not sure about goes to the review screen.
Can it also read order confirmations from suppliers?
Yes. Those are the same steps the other way round: the confirmation is read and laid beside your purchase order, and a difference in quantity, price or delivery date shows up on screen.
How long does the build take?
If your orders are already in a system I can extend, three to six weeks. If an integration with your ERP has to come first, that is added; you hear the lead time up front.

One hour of looking before any building.

In the AI scan we spend an hour together on one process in your business. Afterwards you get a short plan: what AI gains you there, what it costs and how long building takes. Free, and you are not committed to anything.

Tell me what you want to build.

A half-hour call. You get an honest answer about what it costs, how long it takes and whether building it is smart at all.

0:00 / 0:00