The case, explained simply
National Bank of Greece
Read information from documents so staff do less copying and checking.
What actually happened
Microsoft describes NBG using Azure AI Document Intelligence to sort documents and extract information, including Greek-language documents, for digital banking. The supplier case reports processing around 0.5 seconds per page and roughly 90% accuracy. These figures describe reported system performance, not the time required to finish an entire customer request.
What the evidence does not tell us
The account is published by the technology supplier. A single accuracy percentage does not reveal which fields are wrong or how costly those errors are. Public implementation and correction costs are not supplied.
What a Greek organisation can learn
- Choose one document type, such as supplier invoices. List the five fields staff actually need and who may access them.
- Test clean, poor-quality and unusual documents. Compare each extracted field with a checked answer.
- Keep a person checking high-consequence fields such as account numbers and amounts. Measure total processing time after corrections, not reading speed alone.
Before spending money
Ask how unreadable documents are handled and whether you can export the extracted data. Budget for connection to your accounting system and for staff reviewing exceptions.
Read the original evidence
Microsoft · NBG customer story ↗The practical steps above are AIStrategy’s editorial suggestions. They are not reported actions or promised results from the organisation featured. Photography illustrates the subject unless its caption identifies the actual facility.