Blog
Language
Book a demoLog in
Operations

EDI vs Email Orders: Why Customers Still Send Orders by Email

Why large retailers and wholesalers still send orders by email and PDF next to EDI, what that costs your order desk, and how EDI, portals, template OCR and AI agents compare.

Marcus Storm-Mollard
November 2026
7 min read

Customers still send orders by email next to EDI because EDI is set up one trading partner at a time, and much of the order flow never justifies that setup: new customers, small depots, urgent top-ups, order changes and one-off orders. An AI agent can read those emails as they arrive and write the sales order into your ERP, while EDI keeps carrying the high-volume partners.

TL;DR

  • EDI works well for high-volume partners with stable orders. It does not reach every customer, depot and late change.
  • The orders that stay outside EDI arrive as email bodies, PDFs and spreadsheets, and your order desk retypes them.
  • Portals and template OCR each move the work around. A portal asks the customer to change. Template OCR asks you to build a template per layout.
  • An AI agent reads each email as it arrives and writes the order into the ERP. It does not replace an EDI feed that already works.

Why do customers still send orders by email when they have EDI?

An EDI link is a project. Both sides agree on a message format (EDIFACT ORDERS in much of Europe, ANSI X12 850 in North America), a transport such as a VAN or AS2, and a mapping of the customer’s codes to yours. Someone tests it, and someone maintains it when either side changes an item or a depot. That makes sense for a retailer that sends hundreds of orders a week in the same shape. It makes much less sense for a customer who sends two orders a month.

The email keeps arriving for reasons you will recognise from your own inbox:

  • A new customer starts trading with you this month, and the EDI project is a quarter away.
  • A small depot at a large retailer orders for itself and has no link to the central system.
  • A buyer replies to Tuesday’s confirmation with “same again, plus 20 cases for Thursday”.
  • A promotion needs a one-off order with its own prices and dates, and nobody builds a mapping for a single week.
  • An order changes after the EDI message has gone: a cancelled line, a moved delivery day, a swapped item.

Seen from the customer’s side, email is the fastest channel for the exceptions. It needs no setup, and the buyer can write whatever the situation needs.

What does the email side cost your order desk?

Every one of those emails becomes a person’s task. Somebody opens the message, finds the order in the body or the attachment, works out which customer and which delivery point it is for, translates the customer’s article numbers into yours, and types the lines into the ERP. Then somebody checks the result, because a typed order has typing errors.

The late changes hurt most. A change that arrives the afternoon before delivery needs the original order found, edited and re-confirmed, and the warehouse told. When a change is missed, the error surfaces at the depot, as a short delivery, a surplus pallet or a credit note, and by then it is expensive to fix.

None of this is visible in the EDI numbers, which is why the cost of the email side tends to be underestimated. The orders look small one by one. They add up in people’s days.

Can you push every customer onto EDI?

You can try, and for your largest partners you should. The long tail resists. Each new link costs the customer a project of its own, and many customers will not run one for the volume they send you. Late changes and one-off orders stay outside EDI even for partners that have it, because the standard messages are built for the regular order and the exceptions go around them.

So the realistic goal is a split. EDI carries the stable, high-volume flows. Something else handles the rest.

How do EDI, a portal, template OCR and an AI agent compare?

ApproachWho has to changeSetup per customerHandles free-text changesHandles new layouts
EDICustomer and supplierA project: message format, transport and code mappingNo: the standard message carries the regular orderNo: a new format means a new mapping
Customer portalCustomer, who must key orders into your portalAn account and training for each customerOnly in a comment field that someone still readsNot applicable: the portal fixes the layout
Template OCRSupplier, who builds and maintains templatesOne template per customer layoutNo: it extracts fields from a known pageNo: a new layout needs a new template
AI agentNeither: customers keep sending emailYour master data and rules, set up onceYes: it reads “same again, plus 20 cases” against the earlier orderYes: it reads the page, so a new layout needs no template

A portal can work when you hold enough weight with customers to ask them to change. Most suppliers selling to large retailers do not. Template OCR works while the layouts hold still, and customers change them without warning. We cover that difference in more depth in AI agents vs RPA for order processing.

What does an AI agent do with an order email?

The agent watches the order inbox. For each message it decides whether the mail is a new order, a change, a cancellation or something else. It matches the sender and the delivery point to the customer and ship-to records in your ERP, maps every line to your item numbers and units, checks your rules (cut-off times, delivery days, duplicate customer order numbers), and writes the sales order. The step-by-step version is in how to automate sales order entry from customer emails.

It runs on its own. When a field cannot be resolved, the agent holds that one order and sends it to your order desk with the reason, and the rest of the inbox carries on. At a global fresh-produce company, in their first seven weeks, our agents entered 1,000+ retailer orders, about 650,000 cases (about 11,000 tonnes) of fruit, for 63 distribution centres.

Where does the agent stop?

The agent does not replace an EDI feed that works. If a partner sends clean ORDERS or 850 messages, leave that link in place, because a structured message is the cheapest and most reliable input you can get. The agent also does not negotiate with the customer. It enters what the customer ordered under your rules. A request to move a price or accept a late order goes to a person.

The ERP connection is approved by IT once. The agent writes only there, and every order links back to the email it came from, so a disputed line can be traced to its source.

How do you decide which orders go where?

  1. List your customers by order volume and by how regular their orders are.
  2. Keep, or build, EDI for the high-volume partners with stable orders.
  3. Send everything else, including every late change, to the order inbox the agent reads.
  4. Run the agent on a few weeks of real order mail and compare its sales orders with what the desk entered by hand, line by line.

If the comparison finds an order the desk entered wrongly, or a change that was never entered, count those first. They are the cases the depot would otherwise have caught.

FAQ

Why do customers still send orders by email when EDI exists?

EDI is set up once per trading partner, and each setup costs the customer and the supplier time. New customers, small depots, urgent top-ups, order changes, promotions and one-off orders often never get that setup, so they arrive as email and PDF.

Should we push every customer onto EDI?

No. EDI stays the right choice for high-volume partners whose orders follow a stable pattern. For the long tail of customers and for late changes, the cost of a new EDI link is larger than the saving, and the orders keep coming by email.

Does an AI agent replace our EDI feed?

No. The agent leaves a working EDI feed alone. It covers the orders that do not come through EDI: it reads each email as it arrives, matches the lines to your item numbers, and writes the sales order into your ERP.

What happens when the agent cannot read an order?

The agent holds that one order and sends it to your order desk with the reason, for example a delivery point it cannot match. All other orders continue without waiting.

Where to go next

Read how to automate sales order entry from customer emails and how an agent writes sales orders into Business Central. For the wider picture, see AI agents for back-office operations and how Atlas works. Or book a pilot with your own order inbox.

Explore more from Clarm

Helpful links to the product, demo, and policies - all in one place.

Get new Clarm articles

Join the monthly roundup on governed AI agents, autonomous workflows, and source-cited operations.

Talk to us or join the launch list

Book a call to walk through your workflow, or leave your email – we notify you when Atlas opens for new teams.