AI tools now read food orders sent by email

Two in three EU food makers took no electronic orders in 2024, according to Eurostat's survey of enterprises with 10 or more employees. Only 34% received orders through a website, an app, an online marketplace or EDI messages, the structured messages that one company's system sends straight to another's. The rest of their orders came in by other routes, such as a phone call, a typed email or a message on a phone.

Those routes are what a group of software tools for food distributors now read. They take an order written in an email, a text or a voicemail and turn it into an entry in the distributor's own system, so that staff no longer type it in.

Orders typed into an email do not count as e-commerce

Eurostat counts an order as e-commerce when it arrives through a channel designed for placing orders, even if payment and delivery happen offline, and it excludes "orders via manually typed e-mails". When a restaurant sends a list of products to its supplier in an email, or reads it into a voicemail, that order is therefore outside the e-commerce figures, even though the order reaches the supplier through a computer or a phone.

For a supplier, such an order has no fixed format. Product names may be written in the customer's own words, quantities may be in cases, kilos or pieces, and the delivery date may sit in a separate sentence. Someone has to read it, match each line to an item in the price list and enter it before the goods can be picked.

The tools turn a message into an order in the distributor's system

One software company for food distributors says that since 2023 its AI tool has turned orders received by voicemail, text message, WhatsApp and email into digital orders. In March 2025 it launched a further step that, according to the company, sends orders to the distributor's ERP, the software that runs stock, invoicing and deliveries, without anyone entering them by hand. The tool checks each incoming order for accuracy and either processes it at once or flags it for review, and it learns from past order patterns. The company says the tool already handled half of the orders of its early users, without giving their number.

A US start-up building software for food distributors said in March 2026, when it raised $5.8 million in seed funding, that its tools sit on top of the distributor's existing systems and cover order intake, purchasing, inventory planning, invoicing and collections. It said one distributor had freed about 40% of the daily working time of eight sales staff by automating the intake of orders sent by text and email. Both figures are the companies' own and have not been checked independently.

The distributor decides which orders go through without a person

In the first company's description, the approval step stays with the distributor, which chooses which customers' orders go through automatically and which still need manual review, and any order the tool is unsure about goes to a person. The person no longer types the order but checks the cases the tool sends back.

The same split appears in payments, where a pilot launched on 20 October lets agents pay only invoices that a person has already approved. In both cases the software carries out the routine step, and a person approves the decision or the exception.

Greece has the lowest share of food makers receiving EDI orders

In 2024 only 1.6% of Greek food makers received EDI orders, the lowest share among the member states with published figures, against 17% across the EU, according to Eurostat. The shares were also low in Cyprus, at 4.5%, and Malta, at 4.4%, while in Spain 27% of food makers received EDI orders. In Greece, 14.5% of food makers took any electronic orders at all.

Olive oil mills are counted within food manufacturing, and the survey does not publish them separately. Most Greek food makers therefore receive their orders by routes other than EDI or the web, which include the email, phone and message orders these tools are built to read.

Four points a distributor or supplier can check

Which orders: which customers' orders could go through without a person, and which should always be checked, such as new customers, large orders or unusual products.
Product match: how the tool matches the customer's wording to the price list, and what happens when a product name or unit does not match.
Exceptions: who receives the orders the tool flags, and how quickly they are handled before the picking deadline.
Records: the original message kept next to the order the tool created, so that a dispute can be checked against what the customer sent. In freight booking too, the approval step sits in a different place on each system.