Email order processing
From the customer's email to a draft order, with stock and credit checked, pending approval
For companies that receive orders by email and enter them in the ERP by hand, where a missed check lets through a misread item code, stock already committed or credit already used up.
- Data sources
- Email · CRM (the customer management system) · ERP (the company's management system)
- CRM (the customer management system)
- ERP (the company's management system)
- Approval
- Customer Service approves the order; Purchasing approves the purchase order
- First step
- A test with order emails already entered in the ERP, before any new email is processed
Draft customer order
In preparationPending approval- Customer and items identified
- Order within the customer's credit limit
- Stock available; one item below minimum stock
- Draft customer order prepared with every order line
Email order processing turns a customer's email into a checked draft order, ready to be entered in the ERP (the company's management system), with the customer identified, the items matched to the company's own item codes, stock and credit checked and, where needed, a replenishment purchase order prepared. Entering the order in the ERP requires approval by Customer Service; sending the purchase order requires approval by Purchasing.
On this page
- The manual process
- Data sources and the company map
- The six steps for every email
- Reading the email
- Identifying the customer
- Matching the items
- Calculating availability
- Checking credit
- Preparing the replenishment order
- Composing the draft order
- Approval and order entry
- Company decisions
- Rollout
- Limits of the use case
The manual process
In business‑to‑business sales, some orders arrive by email as free text, with each customer's own item codes and often with attachments in whatever format the customer uses. Processing each email by hand involves five tasks across the CRM (the customer management system) and the ERP: identifying the sender in the CRM, locating the items in the catalogue, checking stock, checking credit against the customer's account and entering the order in the ERP.
The main cost of the manual process is not keying time but a missed check, which leaves an order in the ERP with a wrong item code, stock that is not available or a customer over the credit limit.
In the manual process, replenishing an item that the order leaves below minimum stock is handled separately, once the shortage is noticed.
Data sources and the company map
The agent is the software that performs each task. The agent works from the company map, the written description of where each piece of data is held, what each source contains and how each business situation is analysed.
Show the figure
Map contents for this use case
Show details
For this use case, the map lists three sources: the orders mailbox, which holds the order; the CRM, which holds the customer record with the price list that applies to them, their payment terms and agreed credit limit; and the ERP, which holds the catalogue and its prices, stock, purchase orders pending receipt and credit already used. Using those three sources, the map defines four checks and two operations subject to approval.
The agent reads a copy of the data from the three sources, and every figure in the draft carries the date of its last update; if a source does not respond, its part of the draft is marked as out of date and the order goes ahead. The agent never writes to the sources on its own: it writes in two operations only, entering the order and sending the purchase order, and both run after approval. Customer Service approves the order and Purchasing approves the purchase order.
The six steps for every email
Every email goes through six steps, always in the same order: reading the email, identifying the customer, matching the items, calculating availability, checking credit and preparing the replenishment order, the purchase order the agent proposes when an item falls below minimum stock. No check stops the order: each check records its result on the line and the draft goes for approval in full, with every outstanding query flagged. Each step has its own section below, covering what the agent reads and what it records on the line; after the six steps come the draft order, which brings the six results together, and the approval, after which the agent enters the order in the ERP.
Reading the email
Reading the email means extracting the order lines from the email body, its tables and its attachments. The language model (the artificial‑intelligence program that reads text) extracts each line with the customer's original text and its quantity; a line with illegible or ambiguous text is flagged as an outstanding query and left incomplete.
The language model receives only the email and its attachments, with no data from the ERP or the CRM, and does not match item codes: matching is the next step, with the cross‑reference table.
Show the figure
From the email to the order lines
Customer's email
- Line with the customer's item code and the quantity
- Line with the customer's description and the quantity
- Line with ambiguous text
Extracted lines
Identifying the customer
Identifying the customer means looking up the email's sender in the CRM (the customer management system). The agent compares the sender's email address and domain with the CRM records and takes the customer's price list, the payment terms and the agreed credit limit from the record it finds. If the sender has no record, the order goes ahead with an outstanding query on the customer.
A new contact at a company that already has a record is identified by the company's email domain; in this use case, the agent neither creates nor changes CRM records.
Show the figure
The sender, looked up in the CRM
Email sender
- Email address
- Company domain
- Contact name
Customer record in the CRM
- Company and contact
- Price list
- Payment terms
- Customer identified, with price list and payment terms
- No record in the CRM: outstanding query on the customer in the draft
Matching the items
Matching the items means converting each line's item code into the company's own catalogue item. The agent looks up the customer's item code in the cross‑reference table (the table that links each customer's item codes to the company's own) and takes the company's own item and its price on the customer's price list from the ERP. An item code the table does not contain stays as an outstanding query with the original text; the agent does not replace it with the closest match.
The cross‑reference table grows with every approved correction: an item code corrected on one order is matched automatically on the same customer's next order.
Show the figure
The customer's item code, matched to the company's item
Order line
- Customer's item code
- Customer's description
- Quantity
Cross-reference table and ERP
- Customer's item code and the company's item code
- The company's item code and description
- Price on the customer's price list
- The company's item code and price on the line
- No cross-reference: outstanding query with the customer's original text
Calculating availability
Availability is the quantity of each item the company can deliver today, calculated from the stock in the ERP. The agent subtracts the quantity committed to other orders from stock and takes account of purchase orders pending receipt and their dates; the result on the line is the quantity that can be delivered today and the date on which the rest can be delivered.
The ERP's stock field on its own is not availability: it includes stock already committed to other orders. Reading that field as availability leads to confirming orders that cannot be delivered.
Show the figure
The availability calculation
- Quantity that can be delivered today
- Date on which the rest can be delivered
Checking credit
The credit check compares the order amount with the customer's available credit. The agent subtracts the credit used and the unpaid invoices from the agreed credit limit and compares the result with the amount of the draft order. An order above the available credit is not stopped: it goes for approval with a credit flag and the amount by which it exceeds the available credit.
The decision on an order with a credit flag rests with Customer Service. If the company sets a rule for those orders, the rule is written into the map and recorded in the approval log.
Show the figure
The credit check
- Order within the available credit
- Order above the available credit: credit flag with the amount by which it exceeds the available credit
Preparing the replenishment order
The replenishment order is the purchase order to the preferred supplier that the agent prepares when an item's stock, after the order, falls below its reorder point (the stock level that triggers a new purchase order). The agent takes the item's minimum stock, reorder point and preferred supplier from the ERP; without that data, the agent flags the item for review and prepares no purchase order.
Sending the purchase order requires approval by Purchasing, separate from the approval of the order by Customer Service.
Show the figure
From the resulting stock to the purchase order
- Purchase order to the preferred supplier, awaiting approval by Purchasing
- No reorder point or no preferred supplier: item flagged as below minimum, no purchase order
Composing the draft order
The draft order is the document that gathers the result of the six steps: the customer identified with their price list and payment terms, each line with its own item, its deliverable quantity and its date, the outstanding queries flagged and the credit flag where it applies. The draft is complete even with outstanding queries; no check stops it.
The draft is what Customer Service sees on opening the approval request: the whole order, with every outstanding query on its line and the reason for every flag.
Show the figure
The complete draft order
Draft customer order
Pending approval · Customer ServiceApproval and order entry
The draft order reaches the Customer Service inbox as an ‘Approval · order’ request with three actions: approve, change and reject. On approval, the agent enters the order in the ERP as it stands; a change opens the lines for editing before entry; a rejection keeps the order out of the ERP and records the reason.
Show the figure
The approval request
Customer Service inbox
Approval · order
- Draft order with the result of every check
- Outstanding queries and credit flag, line by line
- Replenishment purchase order, where it applies, for approval by Purchasing
- Approval log: the decision and who took it
- Operations log: what the agent read and calculated
Show details
An order with no decision stays pending and is not entered in the ERP. If stock or credit changes between the draft and the approval, the agent updates the draft and requests approval again; an approval is never applied twice.
Every decision is kept in the approval log, alongside the operations log (what the agent read and calculated).
The cycle is always the same: the agent prepares, a person approves, changes or rejects, and the agent enters the approved order. The next order benefits from the corrections: corrected cross‑references are kept in the cross‑reference table, and credit rules are written in the map.
Company decisions
Three decisions specific to each company determine what the agent proposes.
Scope of approval
In this use case, every order requires approval. If the company decides that orders from customers within their credit limit and with no outstanding queries are entered without going through the inbox, that rule is written into the map and recorded in the approval log; the agent does not set that rule itself.
Minimum stock and preferred supplier
The agent proposes a purchase order only if the ERP holds the item's minimum stock, reorder point and preferred supplier. Without that data, the agent flags the item for review and does not prepare a purchase order.
Language model and hosting region
The language model that extracts the order from the email, its provider and the region where it runs are chosen during rollout. The model receives only the email and its attachments, with no data from the ERP or the CRM; the company signs the contract with the model provider.
Rollout
The rollout begins with building the map of order, customer and stock data and testing the agent on emails whose orders have already been entered, before the agent processes any new email. Hellomatik builds the map from the data the company supplies; the company neither accesses the technical console nor writes the map. The use case has no fixed timetable.
- 01
Company map
Hellomatik defines the use case's sources: the mailbox, the CRM and the ERP data for customers, items, stock, orders, purchase orders and credit. The ERP connection reads data without approval and writes only approved orders.
- 02
Cross-references and minimum stock
Hellomatik loads each customer's item codes from their previous orders, and each item's minimum stock, reorder point and preferred supplier from the ERP. Missing data stays as an outstanding query until the company supplies it.
- 03
Testing with orders already entered
The agent processes emails for orders already entered by hand, and each draft is compared line by line with the original order in the ERP. Every difference is reviewed before a new email is processed.
- 04
Inbox and approvals
The company assigns who approves orders in Customer Service, who approves purchase orders in Purchasing, and which inbox each approval request goes to. Once the approvers are assigned, the agent starts reading the orders mailbox.
Limits of the use case
The use case does not say how many orders a company receives by email or how long each one takes to process. No company is running this use case yet: volume and processing time are measured on the emails of the company that adopts the use case.
The use case does not claim every format is read equally well. An order dictated in a voice note or a photographed handwritten order falls outside the use case: the whole order is flagged as an outstanding query, for manual entry. In the formats that are read, data not recognised with confidence is flagged as an outstanding query.
Book a meeting with a Hellomatik engineer to review how email orders reach your company today.
Or request a meeting here

