Yes, AI can connect a customer description to an item without an exact item number in the email. It needs your item master and relevant context, such as previous orders. Ambiguity remains a reason for review. At Voortman, ENTR combines fuzzy matching and order history to connect raw customer emails to intended items in SAP.
Why does item research take so long?
Customers use their own names, older numbers or shortened descriptions, while the ERP uses different terms. Sales support must connect these before entering the order. OCR that returns the literal text does not automatically perform this research.
What is fuzzy matching in order entry?
Fuzzy matching looks for similarities when wording does not match exactly. It can help with spelling differences and abbreviations. It does not make technically different parts interchangeable. Dimensions, variants and packaging still determine the correct choice.
| Situation | Text recognition alone | Business-context matching |
|---|---|---|
| Abbreviation | Return the abbreviation | Compare with item descriptions |
| Older number | Copy the number | Check known item relationships |
| Previous purchase | Read the reference | Use order history as context |
| Similar items | Recognize the name | Consider attributes and expose uncertainty |
| Unit | Copy the quantity | Check unit and packaging |
When is order history useful?
A customer may request the same part as last time. History narrows the search but does not make missing specifications certain. The customer might now use a different machine. Treat a historical match as supporting context, not permission to guess a technical variant.
What does the Voortman case demonstrate?
Voortman received orders and quote requests as free-text emails. Staff manually investigated the intended items. ENTR combines fuzzy matching with historical orders and processes the result in SAP. This example supports the use of context for item matching; it does not prove every description can be resolved without review. See Voortman’s workflow.
How should you test matching quality?
Include a correct product name, an abbreviation, an old number and two items differing only in size. Add an impossible-to-resolve description. The solution should request clarification or require review rather than silently select an item. These are proposed test cases, not customer records.
What changes for sales support?
The value lies in less repeated research and focused attention on unclear cases. Your team need not search the whole catalog for every request. Combine matching with the complete email-to-ERP flow so the selected item does not still require manual entry.
