The “National Bank
with a local accent”
takes on Tech
Evolution with DocDigitizer

Crédito Agrícola is a Portuguese cooperative financial group.

Its genesis is based on the agricultural sector, but its evolution transformed it into a financial group and universal insurance company with a practice of physical proximity and strong relationship with its clients, translating into a strong and trustworthy brand.

Its specific characteristics of decentralization, through Caixas Agricolas (local banks), led to a network of over 620 branches that operate in local markets, actively contributing to counteract the attraction of the populations to the coastal areas, with a dual capacity to provide services and generate employment.

Crédito Agricola’s social responsibility goes way beyond its business motivations.

The Group's liquidity and solvency, together with the high level of trust, are factors that Crédito Agricola’s preserves and constitute a main daily reference to guarantee stability and quality of service.

In 2018 Crédito Agricola started a digital transformation process from a fully manual document process to a Robot Process Automated, one that would bring the extracted information straight into the ERP optimizing processes and streamlining the administrative operations. To achieve this goal the first step was selecting an information extracting solution to integrate the digitalization with the RPA.

The multiple document source, different sizes, formats and content was something that most OCR applications struggled with presenting an insufficient accuracy leading to trust issues in the process.

The team looked at the market for solutions and found DocDigitizer with an accuracy promise of 99,9%

DocDigitizer’s promise is so high that it is almost too good to be true. So the team setup Proof Of Concept with Billing from the central office performing multiple information extraction tests with structured easy to read documents and others that had different structures and were hard to read, even for a human. DocDigitizer’s team was present throughout the whole process providing the necessary support.

The results were surprisingly true to the promise. Receiving multiple documents from multiple sources with different layouts the results were 99.9% true to the document.

No other tested solution came even close to the results presented by DocDigitizer’s AI process with the Human in the Loop process. The error is marginal and not relevant in the whole process of extracting billing information to feed the RPA.

The tested items, among others were:

• Integration speed and accuracy
• Adaptability to new document sources
• Speed of Processing
• Precision of extracted information
• Information Transaction Security
• Customer Support Speed and Efficiency
• Cost of operation

As DocDigitizer registers more usage it learns and keeps improving its accuracy and speed.

Today the process works in 24h document batches, that are processed and the results returned in under 24h so that processing cycles don’t overlap.

The next level is to scale up and reach the whole network and expand the information extraction to more data fields for a more detailed accounting process.