Next-generation contract intelligence.
Evisort uses Adobe PDF Extract API to bring visibility and intelligence to contracts.
Delivers strong foundation for more accurate AI and machine learning models
Objectives
Create central database of contract information to share across internal systems
Ingest PDF contracts with speed and accuracy
Gain more data from contracts using powerful Natural Language Processing models
Results
Provides comprehensive PDF processing for text, tables, images, formatting, and structure
Maintains contract structure to easily identify paragraphs and clauses
Accurately identifies table and cells from PDF
Adobe PDF Extract API delivers accurate content extraction with better results, providing better input for deep learning models
When Jerry Ting graduated from Harvard Law School, he was excited to start advising clients on strategy and helping them solve their business problems. Instead, he realized that much of his time was spent manually reviewing contracts and other legal paperwork.
One of Ting’s first legal clients was Amine Anoun, then a graduate student at MIT working on advanced artificial intelligence research. Seeing Anoun’s work sparked an idea: What if they could use AI to automate all of the mundane aspects of contract work? If AI could minimize tasks such as reviewing contracts, lawyers could spend more time advising their clients. Together with Harvard classmate Jake Sussman, the three founded Evisort in 2016.
Evisort offers a completely new type of contract management platform. Using advanced machine learning algorithms, Evisort provides an end-to-end central platform for contract management and analysis. Evisort understands the meaning of words inside of contracts, without needing humans to manually do tedious data entry, freeing them up to be more strategic with their time. Key data points like party names, payment terms, expiration dates, and legal obligations are tracked automatically. Companies can use the platform to build contracts based on templates, automate review cycles, store contracts in a central repository, and analyze terms.
“The average company loses 10% of a contract’s value due to mismanagement,” says Ting, now CEO of Evisort. “This could be anything from missing out on discounts for early payment to being locked accidentally into a new term with a vendor by missing an expiration date and automatic renewal. By providing greater visibility into contracts, we help companies reduce risk and move contracts through the pipeline up to 46% faster.”
Evisort improved its ability to read and understand contracts by incorporating Adobe PDF Extract API into its intelligent contract platform. Working with PDF Extract API, Evisort can ingest contracts with greater ease, leading to faster and highly accurate data for clients.
“Adobe PDF Extract API is not your typical OCR solution. It looks at documents at the holistic level, the same way that a human sees it. Text, formatting, tables, images, indentation — everything is preserved.”
Amine Anoun
Co-founder and CTO, Evisort
Detecting structure from PDF contracts
One of the most sought-after functions of the Evisort platform is its ability to ingest existing contracts and extract all relevant information into a central contract repository. The repository then syncs data with ERP, CRM, and other key systems across a company, aligning expectations and streamlining workflows. Anoun, now CTO of Evisort, and the rest of the Evisort AI Labs team use the Java SDK to bring greater speed and accuracy to the platform’s contract management capabilities with PDF Extract API.
Evisort can ingest data from contracts that are saved as PDF files or even as e-signature files. While the Evisort AI algorithms are second to none, results are highly dependent on the system’s ability to read and interpret contracts accurately. Anoun tried using optical character recognition (OCR) solutions from leaders in the field, but even the best OCR solutions often fell short, especially on scanned images with tables like invoices.
In the end, Evisort wanted to work with a comprehensive, easy-to-use solution that would provide all needed features out-of-the-box. This would give developers more time to focus on training the AI and refining core capabilities of the Evisort platform. Anoun turned to a company that he knew had expertise in working with PDF and electronic documents: Adobe. Only PDF Extract API provides the level of structural and stylistic information that Evisort needs to accurately break contracts into paragraphs, clauses, sub-clauses.
“Adobe PDF Extract API is not your typical OCR solution,” says Anoun. “It looks at documents at the holistic level, the same way that a human sees it. Text, formatting, tables, images, indentation — everything is preserved.”
With an accurate understanding of the document structure, Evisort algorithms can learn to extract party names, expiration dates, company obligations, and more than 60 other individual points of metadata with greater accuracy and speed. PDF Extract API even identifies the language at the text block level, instead of at a document level. This is particularly helpful when parsing contracts between international parties where paragraphs may alternate between the English text and the translation in another language.
For Evisort, the ability of PDF Extract API to accurately extract tables into JSON or CSV format has also been groundbreaking.
“If someone is looking for a table extraction tool, I would recommend starting with Adobe PDF Extract API,” says Anoun. “There’s nothing else on its level.”
Typical OCR solutions often have trouble finding the edges of cells, resulting in unorganized text groupings — exactly the opposite of what Evisort needs from tables. PDF Extract API maintains cell structure, allowing Evisort to accurately parse critical information such as fee schedules or information about parties in a contract.
“Adobe PDF Extract API is at the core of what we do because it gives us highly readable data so that our algorithms can pull data out more efficiently. It’s exciting to see the impact that AI brings to the world of contract analytics and management.”
Jerry Ting
Co-founder and CEO, Evisort
Bringing AI to contracts across industries
More than 100 companies manage contracts with the Evisort platform, including Microsoft, Keller Williams, and BNY Mellon. Many contracts touch multiple teams — legal, procurement, sales, and finance. The Evisort platform becomes a single record of truth that feeds data into all other company systems to remind procurement about upcoming renewal dates, alert finance to payment deadlines, or deliver a record of obligations to each department involved in services. Microsoft uses to use the Evisort platform to analyze thousands of contacts every year, eliminating hours of manual data entry and reducing the risk that critical contract terms might slip through the cracks.
While Evisort is able to customize machine learning models in addition to out-of-box intelligence offerings today, the company aims to create more self-service ways for customer to work with Evisort technologies and create truly custom algorithms to meet specific needs with very little training data. PDF Extract API will provide a critical foundation for the self-service model by providing users a more accurate structure that they can build their deep learning models upon.
“Our customers work with us because we can eliminate time, cost, and risk associated with having to manually record critical information into contract management systems,” says Ting. “Adobe PDF Extract API is at the core of what we do because it gives us highly readable data so that our algorithms can pull data out more efficiently. It’s exciting to see the impact that AI brings to the world of contract analytics and management.”