Google Document AI for Contracts & Resources Management

May 25, 2023
Posted by Oksana Gladkikh

Business Need

Companies deal with documents that contain critical data to maintain business operations on a daily basis. These routine activities may take plenty of time of highly qualified staff. The manual labor to process these documents is tedious and error-prone - also adds to document processing cost and time.

To overcome these difficulties, the following goals were formulated:

Accelerate and simplify contract life cycle management by digitizing and extracting highly accurate data from contracts
Ensure processing of the documents can be adjusted while new contract structures are introduced
Reduce manual document processing, minimize setup costs, and accelerate deployment
Ensure a high level of accuracy with Human-in-the-Loop (HITL) reviews

Akvelon’s Engagement

Akvelon’s team started with investigating the service providers which can be seamlessly integrated with Contracts & Resource Management system developed by Akvelon.
Akvelon Solution Architects chose Google Document AI for the following reasons:

Document AI platform already contains pre-trained models for document processing, including basic extractors like OCR and Form Parser
Providing Document AI Workbench to train the models on different business documents or create own models to get better results for documents
Presence of Human-in-the-Loop AI to help achieving higher document processing accuracy with the assurance of human review
Solution should work well with the on-premise deployed software ensuring the highest level of security

As a result, developers created a microservice which communicates with Google Document AI via HTTPS connection and extended current back-end services to integrate workflows of bi-directional communication between Google Document AI cloud and Contracts & Resource Management software.

The parsing process starts from the raw document and moves through our workflow to extract high-value information by converting the unstructured data into structured data.

Along with Google Document AI integration, Contracts & Resource Management software workflow has a user-friendly interface so that document and extracted data can be interfaced with the existing business processes.

Newly received contract PDF is imported into the Contracts Management system.

After that the system automatically processes a contract using Google Document AI to parse the contract and display pre-filled import form using Human-in-the-loop approach for final confirmation of contract terms like dates, invoicing milestones, service terms entries and more by the accountant user like in the example below.

Due to more than 92% accuracy in automated parsing, contract import process now takes just a few moments to complete.

Business Impact

Key impacts since the solution was released to production environment:

Achieved 125% improvement in data freshness/availability in the contract tracking system
Saved about 1.5 months of finance professionals annually reducing mundane data entry work
Supporting more than 20+ contract and purchase order types (SOW, PO, Amendments)
Implemented secure machine learning business solution using Google Cloud Document AI technology, achieving 92% accuracy in contract data recognition

Technologies used:
Contracts & Resource Management system: React JS, .NET Core, MS SQL RDBMS
DocAI Microrservice: Python, Google Document AI API, openCV, Pandas, Natural Language processing library for Python (NLTK), Numpy