Why does order processing in wholesale always get stuck?
In wholesale, you rarely deal with a single type of customer or a single type of order. Retail customers send a purchase order as a PDF. Foodservice parties use their own procurement portal. Smaller buyers email a list in the body of the message itself, without an attachment. And a handful of customers once had an EDI connection built that now half-works because the format on their side has changed. All that incoming traffic lands at one point: the internal team. They re-enter the order lines into the ERP, check whether article numbers match their own references, follow up when a quantity is missing, and send a confirmation back. Each order easily takes ten to fifteen minutes. At a hundred orders per week, that amounts to a part-time role that exists solely for re-entry.
The mismatch problem: customer reference versus internal article code
The real pain point in wholesale is not the re-entry itself, but the translation between the customer reference and the internal article code. A customer orders 'House Brand Fruit Drink 1L, art. 884421'. Internally, that article is called 'FD-1000-HM' and sits under a different product group. That translation lives in no one's head, or in a spreadsheet maintained by the most experienced employee. If that person is out sick, processing stalls or errors end up in the order confirmation. Systems such as SAP or Dynamics 365 Business Central can store customer-specific article references, but that only works if the mapping is properly maintained. In practice, that table is outdated, incomplete, or simply never set up.
What does this actually cost wholesale businesses?
The direct costs are labour hours. But the hidden costs are greater: errors in order confirmations lead to returns, credit notes, and helpdesk queries. An order processed too late causes a missed delivery window and a penalty from the retailer. And the internal team that spends the day re-entering data never gets to customer contact, upselling, or resolving complaints. These are costs that never get calculated, but every operations manager recognises them when they stop to think about it.
When does automating order processing make sense?
Automation only makes sense when volume is high enough and the process is reasonably stable. For a wholesale business processing eighty to a hundred orders per week from a fixed customer base, the business case is straightforward: the repetition is there, the time investment per order is measurable, and the error risk is known. It makes less sense when orders vary significantly in structure, when customer data is structurally incomplete, or when the ERP is so outdated that automated delivery is technically not feasible. That last point is an integration question, not a document AI question. dottle reads the incoming document, converts it into structured order data, and delivers it to the system. A person approves the order before it is processed: deviations are flagged, not silently posted.
What is a realistic starting point for wholesale?
Start with the ten customers that together generate the most order volume. These are also the customers whose formats are most familiar and whose article references are most completely recorded in the system. Once automation runs reliably there, expand. Not the other way around: starting with the most complex incoming traffic always leads to disappointment and a project that stalls. A stable process for the core of the customer base delivers more time savings than a half-working solution across the full range. And it gives the internal team the space to treat exceptions as actual exceptions, rather than the norm.