15 min → 30 s
per order in the real case linked below, from manual reading to entry in SAP
Order automation · AI + ERP
Orders entered into the ERP automatically. No typing.
Email, PDF, Excel or portal: each order is read, validated against customer, item and price-list master data, and recorded in your ERP. Your team reviews only mismatches.
No templates per client: AI interprets the format of each order.
With your ERP: SAP, Business Central, Business One, Odoo, Sage, A3 or, if no other route is available, through the screen interface.
A pilot using real orders, operational in four to six weeks.
4 channels
email, PDF attachment, Excel and customer portal, each with its format
3 master datasets
customers, items and price lists: every line is validated against them before any change is made in the ERP
Exceptions only
prices outside the price list, unknown references or quantities outside agreed terms are sent to a person
The order is already written. The customer wrote it, and someone types it all over again.
Automating order entry means reading an order received by email, PDF, Excel or portal, validating it against customer, item and price-list master data and recording it in the ERP without anyone typing it. A person intervenes only when something does not match.
An attached PDF, an Excel file with columns in a different order, an email with line items in the message body, a customer portal that needs checking. Each channel and customer uses a different format, so order entry becomes manual work: read, find the reference, check the price list, type it into the ERP and hope there are no mistakes.
During demand peaks, orders wait hours before appearing in the system. Errors in references, quantities or prices are discovered late: in the warehouse, on the invoice or through a complaint.
Akaen converts any incoming order into a validated order in your ERP. AI reads the document without templates, every data point is checked against your master data, and only mismatches reach a person. The original remains linked to the record so you can always refer back to it.
How it works
How order entry automation works: from the inbox to the ERP, with an exception resolved by a person.
Example data. The order includes a price outside the price list, so it is sent to sales instead of being entered automatically.
With your ERP
Integration with your ERP: it connects to the system you already have, without replacing it.
The entry route depends on what your ERP supports. We check it in the first week, before building anything.
| ERP | How the order enters |
|---|---|
| SAP (ECC, S/4HANA) | BAPI, IDoc or REST API. The order is created as if it had been typed by a user, with its log. |
| Microsoft Dynamics 365 Business Central / Navision | OData API or web services. Master data validation is included in the same call. |
| SAP Business One | Service Layer or DI API to create the sales document with its lines. |
| Odoo, Sage, A3 and other ERP systems for SMEs | An API where available; otherwise, file import or automation through the interface. |
| ERP without API or import | RPA: A robot inserts the validated order into the ERP screen, with traceability. |
What happens when something does not match
Exceptions are part of the design, not a failure.
A system that only works with perfect orders is of little use. Its value lies in how it handles imperfect ones: detecting them before they affect the ERP and sending them to the right person with the decision ready to make.
Price outside the price list
The order shows €18.40 while the current price list says €17.20. Nothing is changed in the ERP: sales decides in a review queue whether to accept the order price or apply the price list.
Non-existent reference
An old reference, free-text description or customer code. The system proposes the most likely match, a person confirms it, and the mapping is retained for next time.
Quantity outside agreed terms
An incorrect case multiple or a quantity above the framework agreement is flagged before the order is created, rather than after delivery.
Ambiguous customer or delivery address
Several branches share a trading name: the system asks once and saves the rule.
What your company needs
Four things, and none involves changing your system.
A real sample of 30 to 50 orders, covering the most common formats and the most problematic ones.
Reasonably clean customer, item and price-list master data, or a commitment to clean it as part of the pilot.
A way to create orders in the ERP: API, import or, at worst, the screen interface.
A person from sales or administration to resolve exceptions during the pilot.
When it pays off
And when it is better not to do it.
It pays off when
Several hundred orders per month that today someone types from a PDF, an email or an Excel.
Different formats per client, which make fixed templates useless.
Demand peaks during which orders wait hours before entering the ERP.
Errors in references, quantities or prices discovered in the warehouse or on the invoice.
It does not pay off when
Orders already arrive via EDI or via a portal integrated with ERP.
The volume is low and the formats are homogeneous: a well-designed import is enough.
Master data is so outdated that validation would have no reliable reference to check against.
Estimated impact
Calculate the impact with your numbers.
133 h
capacity released per month
Orders × (minutes today − minutes afterwards) ÷ 60. Freed capacity is not an accounting saving: implementation, maintenance and usage costs must be deducted to calculate the return. We establish this using your data.
Real case
B2B orders in SAP for a food industry leader.
Thousands of outlets and a critical volume of B2B orders received in PDFs and emails with varying formats. An event-driven architecture on Azure, generative AI extraction and validation against SAP master data before automatic entry.
Project led by Víctor Muñoz, now an Akaen partner.
15 min → 30 s
time per order, from manual reading of the PDF to registration validated in SAP
5.000+
monthly orders (estimate) in a leading food and beverage corporation
1,200 h
of capacity freed each month in the order entry team
Azure + AI
event-driven architecture with generative extraction and validation against master data
Frequently Asked Questions
Questions from purchasing, sales and IT.
No. OCR converts an image into text. Here, AI understands the document: its header, line items and quantity column, even when each customer labels it differently. It then validates every data point against your master data, which is where the value lies.
No. Reading does not rely on fixed templates; it interprets each order’s format. Business rules are configured once: price lists, multiples, framework agreements and delivery addresses.
Yes, with SAP, Business Central, Business One, Odoo, Sage or A3, through an API or import. If an ERP has no entry route, a robot enters the validated order through the screen interface. We check this in the first week of the pilot, before building anything.
A pilot using real orders is usually operational in four to six weeks: two weeks to connect the ERP and configure the rules, with the rest spent processing orders alongside the team until the exception rate stabilises.
A fixed price agreed before work starts, based on scope: the number of channels, formats and systems. No per-page licences. To see how it would apply to your company first, request the free first look and select “order entry”.
See what it would be like in my company
Request the first look and select “order entry” and you'll see it applied to your sector.
In less than 72 hours, focused on your documents and your ERP. Free, with no obligation.