Intelligent Document Processing (IDP) - 8 weeks implementation

Nearly all business processes begin, include or end with a document or an email. IDC forecasts that new data creation will grow up to 175 ZB by 2025. Manually extracting data, processing it, and creating documents based on these data can be tedious for employees and result in errors, decreased productivity and potential security and compliance breaches.

The Intelligent Document Processing (IDP) solution from R Systems leverages the power provided by the latest cloud technologies, cognitive services, and machine learning to accelerate the digital transformation processes of an organization.

IDP offers a cost-effective solution for automatic processing of images and scanned or digital documents with a high degree of accuracy compared to traditional manual processes.

You can use IDP to extract data from several types of documents utilized across different industries and scenarios:

• Customer onboarding

o extract data from id cards, passports, driving licenses; these data can include last name, series number, CNP, issuer, issued and expiring date, sex, nationality, county, place of residence, street, number, building, floor, apartment

• Accounts Payable & Receivable

o extract data from invoices, purchase orders; these data can include sender name, supplier ID, tax ID, invoice number, currency, total amount, VAT, products, due date

• Claim Processing & Policy Issuance

o Extract data from documents such as health claims, insurance claims, medical reports, policy documents

• Mortgage Document Processing

o extract data from loan applications, forms, legal letters

Potential benefits of implementing IDP for your company:

• Cost savings: There is no competition in terms of cost effectiveness

• No upfront investments in licenses: IDP works on a pay as you go and pay per page/document model

• Time effectiveness: Cloud technology can scale up immediately, thus heavy-duty processing is always possible (unlike scaling up human teams) and hence obtaining much faster the desired output

• Running round the clock: Most components have zero dependency on hourly shifts or working days. Can be configured to work 24x7

• Reduced dependency on humans: Almost zero human dependent processes (except for QA process). The less dependency on humans always brings higher predictability.

• Performance and Quality: Higher productivity against manual jobs with better quality, when accurate data sources are provided to the learning algorithm

• IDP can be combined with RPA (Robotic Process Automation) to automate end-to-end processes with little to zero human intervention.

The architecture of the solution works as follows:

• A user or service can initiate an API request to extract metadata from a document using a secured and encrypted HTTPS protocol and a security key that works as an authentication method.

• The API service connects to the Azure Cognitive Service using a secondary security key. The OCR (Optical Character Recognition) service from Azure Computer Vision will return an unstructured JSON response to the API. This response will be managed by the API that will restructure the file into an XML file.

• The API service has capabilities to verify the metadata using programmatic ML models. These validity checks have the purpose to filter ineligible responses that may come from the cognitive service. Then, the API services will extract and structure metadate from documents like id cards, invoices and more.

The solution will be deployed in an Azure subscription using the following services:

• Container Registry

• App Service Plan

• Cognitive Services/Computer Vision

• Key Vault

IDP Business Model includes the following processes:

• Analysis Phase

• Specifications document & solution architecture

• Algorithm training and fine-tuning

• Services exposure though APIs

• Data & document validation (business logic)

• Testing/UAT

• Training

• Production Migration & Go-Live

• Project Management

• Training and support

This offer is best suited for companies that need to process around 20k pages of identification documents (id cards, passports, driving liceses) and invoices and it incluses the Azure infrastructure costs.