Jerry Ting, is the Founder and CEO of Evisort a company offering contract workflow and management capabilities to improve efficiency, reduce risk, and gain visibility into any contract. Evisort is the first platform to automatically aggregate contracts from across the business, create streamlined approval processes, and apply advanced artificial intelligence (AI) to provide real-time analysis so your legal, contract, and procurement teams can accelerate your businesses.
What initially attracted you to AI?
I first started thinking about artificial intelligence while working as an undergraduate intern in the U.S. Supreme Court. I noticed how time-pressed lawyers were spending thousands of hours organizing, reviewing, tracking and performing due diligence on legal documents. It was crazy to see the highest court push around bins and bins of paper for evidence. That gave me the idea that AI could make this tedious labor much more efficient, given its ability to quickly parse huge datasets.
You had the idea for Evisort while attending law school. Can you walk us through this genesis story?
I had done a stint working in sales at Yelp prior to law school, where I saw that legal and vendor management teams faced the same problems that lawyers did with reviewing large amounts of data. I realized that if AI could analyze documents in litigation, it could also comprehend complex business contracts. I realized just how cumbersome it was to try to do the legal work with basic software but without automation. In my discussions with professors and alumni of Harvard Law, a common thread was that people were spending a lot of time looking for language, clauses, precedents, dates, and party names in contracts. It really was a manual process.
I teamed up with data scientists from MIT and lawyers from Harvard to create an artificial intelligence platform, Evisort. I met my co-founders Amine Amoun and Jacob Sussman through the Harvard Law Entrepreneurship Project. Jacob was in my class of 2018 at Harvard Law, and Amine was working on AI at MIT. We met Memme Onwudiwe and Riley Hawkins in the 2019 class, and invited them to join the team.
The Harvard Law community helped us get the business off the ground in the early days. We won best presentation at the 2017 Lee Kuan Yew Global Business Plan Competition in the services category. With support from mentors and deans at Harvard Law, we had an opportunity to showcase our software to corporate and public sector lawyers, such as staffers from state attorneys’ general offices. This was immensely helpful in shaping our product-market fit.
Were you at all surprised that no one else was offering a product like Evisort?
Absolutely, I was surprised. I knew even that Microsoft Excel could do things to help attorneys. But why aren’t they doing it? I thought, “I need to go find someone to build the platform because this is a billion-dollar opportunity.” The legaltech space is chock full of startups and established players offering quite sophisticated solutions for ediscovery in litigation, some with very good AI and machine learning. However, there are far fewer products on the other side of the legal house. Transactional attorneys research, draft, and review contracts, handle large corporate mergers and acquisitions, manage securities offerings, and prepare closing documents for the sale and purchase of real property.
The data contained in contracts is not only meaningful to the attorneys who draft, review, and process those documents. These data also drive how finance departments bill and send out invoices, or how, when, and to whom a sales team would sell a given product or service. It even drives what they choose to sell. These data define a given company’s relationships with its suppliers, clients, and employees and, in many ways, underpin and guide day to day decision-making. You can break down a company into its entire value chain by looking at its contracts.
Have you been taken by surprise at how receptive the marketplace has been to Evisort?
Evisort works for teams from startups to Fortune 100s with complex legal and procurement processes. The pandemic of 2020 held the greatest surprises in our business growth. Many lawyers have commented how they did not expect they’d be spending so much time reviewing Force Majeure clauses. The uptick in this work, as well as the distributed nature of teams, has led to the number of new business queries growing by five times the rate we saw in 2019.
Could you walk us through some of the machine learning technologies that are being used at Evisort?
Evisort’s software integrates optical character recognition and natural language processing to extract metadata from the client’s contract, and then to allow the client to easily search through this metadata. After importing a document into the app, the software automatically runs OCR and enhanced legal contextual spell-checking to gather data for machine processing. On top of that, layers of AI will 1) classify the document using metadata; 2) extract and identify relevant information at the paragraph level using a combination of classification methods such as linear models, decision trees, and neural nets; and 3) extract specific entities from the text using recurrent neural net technology, such as long short-term memory (LSTM) and convolutional recurrent neural networks (CRNN).
Evisort is the first authentic AI that can identify contract type, non-standard third-party clauses, counterparty names, expiration dates, and automatically tie parent to child. We have trained these layers of AI on over 10 million documents, such as NDAs, purchase agreements, and other commonly-used types of contracts. This means it can understand over 230 types of contract and doesn’t have to be trained on each individual company’s own contracts. We encourage any prospective user to give us any contract and we can show them that our AI works right in front of their eyes.
Legal contracts are often full of obscure legal terminology where context is everything. What are some of the unique natural language understanding (NLU) challenges that Evisort has had to overcome?
The biggest challenge we have had to solve is around scanned documents. From a NLP perspective, working to analyze legal text is a challenge that we have pioneered and solved. What we did not expect was to have to help redefine the computer-vision field of optical character recognition, which is the technology used to analyze images. Many legal contracts are scanned and actually images, not text, and existing OCR technology has about a 20% error rate. We built new OCR technology that reduced that error rate to a fraction of 20% and it drastically improved the accuracy of our algorithms.
How much time can legal departments save by using this type of technology?
We got our hands on more than 10 million real-world documents and pretrained our AI to recognize NDAs, purchase agreements, what have you, out of the box. That means that our customers in legal and procurement can begin using the software from day 1 of install. The average AI solution available off the shelf often requires seven to ten months of consultant-driven configuration before they can begin uploading live documents from their legal operations into the system. That’s in addition to the cost and time savings reaped by responsive search. The time savings are important, but even more compelling, we hear from customers and prospects how this tool makes them more efficient at their jobs, and frees up time so they can focus on other important jobs that require their attention.
Evisort also offers procurement solutions. Could you discuss some of these solutions?
Both legal and procurement teams are using the same AI-based software we provide. Evisort first converts scanned documents to searchable text—nothing new here. But it’s the next steps that have a revolutionary application for lawyers and purchasers. Using artificial intelligence, Evisort sorts through all the contracts, categorizing them by subject area and type of contract, and identifies provisions within each contract. A whole range of key data is extracted such as party names, dates, and size of the deal.
Let’s say a procurement manager in the middle of finalizing a high-value vendor contract comes to general counsel seeking guidance on negotiating a provision on the limitation of liability. Right now, the lawyer can do a word search for “limitation on liability” among the contracts the legal team has access to in order to find relevant contracts, but she would have to open each one to read it and see if it’s helpful. Our software instantly scans every contract in the entire company that includes a limitation on liability—pulling up only those within a certain date or other parameters that the lawyer wants. It presents this data in chart form. The chart shows when the contract was signed, how much money was involved in the deal, the language of the limitation on liability. This saves the lawyer multiple hours of work reading each document by hand.
Likewise, procurement teams can perform word and date searches to understand exactly who owes them what, and when it’s due. Procurement and contract managers turned to contract lifecycle management tools like Evisort as their businesses started missing auto-renewals, looking to exit leases or seeing revenue impact from shipping delays. We saw a huge surge in this activity in 2020 with the pandemic and related trends.
In July, we launched Evisort Contract Workflow, which enables procurement managers to get their contracts signed more quickly. This uses the data from our Contract Management product to automate the creation of new contracts, using legal-approved templates. A premium risk-identification module identifies problematic passages and suggests amendments based on preset standards. Our tool helps catalyze the most efficient approval process. Procurement managers can assign approvers based on different conditions and terms and review all contracts pending signature at a glance through our dashboard. Together, Contract Workflow and Contract Management, help keep all members of procurement and legal on the same page about contract status and obligations.
Is there anything else that you would like to share about Evisort?
While the primary application of our AI is end to end contract management, we are really an AI company. We used our Series A funding to invest in training our AI so it’s unparalleled in its ability to recognize the use cases customers in a range of different types of business are asking about. We’re continuing to invest in the Evisort AI Labs (our AI R&D team) and explore new ways we can help businesses apply our authentic AI. We’ve hired across every department in 2020, as we see this tremendous surge in leads and interest using our product.
Thank you for the great interview, readers who wish to learn more should visit Evisort.
- Researchers Mimic Sea Slug Strategies in Quantum Material
- Do Conversational Agents Like Alexa Affect How Children Communicate?
- Hobbling Computer Vision Datasets Against Unauthorized Use
- Faisal Ahmed. Co-Founder & CTO at Knockri – Interview Series
- The Shortcomings of Amazon Mechanical Turk May Threaten Natural Language Generation Systems