Overview
The proprietary ZS solution is built on top of the ZS’s Intelligent Document Processing engine, a group of assets that can be leveraged to set up an unstructured ETL pipeline that pre-processes the documents, performs necessary processing and post-processes the extracted data. Businesses across industries are leveraging the enterprise cloud-based offering for the following: • Contract digitization: To perform automated extraction of data from PDF contracts for downstream processing. • Invoice and purchase order digitization: To automatically extract unstructured and semi-structured data from financial documents. • Form digitization: To extract metadata from digital and manual forms and sources, and collate the extracted information in a structured format.
The solution leverages a combination of pre-developed NER API offerings, use-case-driven customer NER models and rule-based extraction to maximize the accuracy of the output. Additionally, for low-confidence or business-critical extractions, the system can set up a human review pipeline leveraging an intuitive role-based user interface for data validation.
ZS Document Automation has enabled organizations to improve efficiency and mitigate risk. Following are some use cases where it was applied: • Expense spend audit to detect frauds: A biopharma company faced challenges in the manual audit of food and beverage invoices. The ZS solution tackled the challenge by extracting data from unstructured receipts and converting the data into a structured format. An AI-driven algorithm then flagged any anomaly in the data, which increased the accuracy of fraud detection. • Invoice digitization to drive reward and loyalty program: A firm collected point of sales invoices, which contain rich information about consumers and accounts but struggled to unlock the value of data source. The document automation offering digitized the invoices and converted the data into a structured format which enabled the firm to gather insights from invoice data to drive value for consumers. • Sales contract digitization to extract metadata and revenue information: An organization’s sales support team used to manually audit and record metadata from thousands of contracts. The ZS solution integrated and validated the intake of contracts, as well as automated extraction of metadata to identify opportunities to improve efficiency. • Consent management: An enterprise had a huge volume of documents, which made the extraction of consent metadata a business-critical challenge. The ZS IDP solution’s document reading capability supported the extraction of consent-related metadata from digital and manual documents, which helped in the reduction of manual effort
Highlights
- The time required to extract and process documents goes down significantly, thereby reducing dependency on FTEs.
- Human-centered design allows human review in the data pipeline, which enables businesses to further improve accuracy and capture data for NL model enhancements and generates an aggregate report for stakeholders.
- Data extraction accuracy with IDP augmentation is as high as 98%, with complete mitigation of human errors. The solution leverages various NL APIs, such as AWS Textract, in conjunction with custom models hosted on highly secure and scalable AWS architecture.