The Evisort Challenge: How to Tell True AI Systems Apart from the Rest
Would you purchase a contract AI system without first testing it live on your own contracts?
When we ask this question, most people say that of course they wouldn’t. However, there’s a vital distinction between testing the system on the vendor’s schedule, and actually testing it live.
Most of the in-house lawyers and procurement managers we talk to say that digitizing the contract management lifecycle is an increasingly important priority. Manually managing contracts using nothing but email and spreadsheets simply isn’t feasible at scale. Businesses need to create fall-back versions of their agreements, automated playbooks to support contract negotiation and management, and more dynamic clause libraries. In addition, business teams are increasingly aware that post-award management of vendor contracts is where they derive the most value.
This shift in the contract management landscape requires innovative thinking and effective change management. Moving to an innovative approach to contract management requires the implementation of an innovative contract management system.
But when you’re comparing different contract management software vendors ”” and especially if you’ve been burned in the past by a vendor that over-promised and under-delivered ”” you probably want to know how to test whether a company genuinely offers cutting-edge artificial intelligence (AI) that will make your life easier.
Before you can do that, however, it’s necessary to examine what AI is, and what you should expect it to do for you today.
What is AI?
Artificial intelligence in contract management refers to the application of machine learning and natural language processing to contract drafting, review, and management.
Machine learning isn’t some sort of alien process. AI algorithms learn the same way we do ”” they just do it at a massive scale that makes them much faster than any team of humans could ever hope to be. Machine learning accomplishes this speed and scale through a combination of active learning and deep learning.
Active learning is like learning a language in a classroom. Your teacher gives you vocabulary lists and teaches you rules of grammar and syntax. With that conceptual structure, you’re ready to start communicating. Deep learning, conversely, is like learning a language through the immersion method. You absorb as much information as you can and you start to make your own connections in order to develop an understanding of the language and start communicating.
Each approach has its benefits. In contract management, active learning means the AI has been pre-trained to recognize dozens of common clauses based on real-world examples. Deep learning enables the AI to continue learning as you put it to work, making its own connections and developing an understanding of the special characteristics of your industry and the unique needs of your business.
Combined with natural language processing capabilities, machine learning empowers the best contract management solutions to quickly and accurately parse mountains of data contained across hundreds of thousands of contracts. That reliability frees legal and business teams to stop spending hundreds of hours doing rote work, and instead focus on the substance of their commercial agreements and the formulation of procurement and contract management strategies.
All AI is not created equal
We’ve looked at what a leading AI platform can do, but not all companies that claim to use AI are really using AI. Truly innovative AI platforms set themselves apart from the rest of the crowd in three key areas: accuracy, precision, and speed.
A key differentiator between AI leaders and the rest of the crowd is accuracy. Many solutions only offer accuracy as low as 60%. If you have thousands of contracts in your portfolio, that means you’ll still have hundreds of hours of menial work in front of you after using subpar AI.
For example, if you have 10,000 documents to review and the contract management system you’re using is able to identify them with only 60% accuracy, then you still have 4,000 documents to review manually. Compare that to an AI system that offers 90% accuracy, which would cut the number of documents you need to review down to 1,000 ”” a 75% decrease. If you’re outsourcing document review to a team that takes 20 minutes per document, the better AI saves you 1,000 hours of review on the 3,000 documents that the first system couldn’t accurately identify. At an hourly rate of $33, that’s $33,000 you just saved for your business by investing in true contract management AI.
In addition, the best AI can identify not only matches, but also subtle differences. When you’re sorting through tens or even hundreds of variations of a clause, you need a system that can recognize and tell you when the language in a particular clause is close, but not exactly identical, to what you’re looking for. By feeding the system your business’s existing documents, you can properly train it on your business’s unique needs. That additional fine tuning takes a system that’s already a highly capable turnkey solution and transforms it into a smart assistant that can tailor its services to your business.
Many contract management systems can identify specific clauses, such as those addressing limitations on liability, termination of the contract, and governing law. However, not all contract management platforms can actually extract the data you need from those clauses, so you’re still left to review them and parse the actual data yourself.
To illustrate, let’s say that a contract provides that the termination date is six years from the start of the contract. The AI needs to be able to identify the effective date of the agreement and then count six years forward in order to identify the termination date for you and save you the time you’d otherwise spend manually looking up and figuring out the dates yourself (which is even trickier when future dates are calculated with respect to business days instead of calendar days).
If you want a platform that can automate data extraction in that way instead of a platform that will just point you in the right direction and require you to do the actual work, then you need the best AI working for you.
This aspect is a function of the first two. The more accurate and precise the AI system is at identifying and sorting the data in your documents, the faster your contract review becomes ”” because you’re not wasting time correcting the work of a subpar contract management system. The faster you can get through more basic tasks such as document review and template creation, the more time you have to focus on more substantive work, from comparing vendors to optimizing your team’s workflows.
Why do you need real AI?
Given the discrepancy between the capabilities of platforms that leverage actual machine learning and those that don’t, some vendors look for other ways to boost their accuracy. When a contract management software provider is working with less sophisticated tools, they might use human intervention in order to supplement their algorithms’ more limited capabilities.
Instead of an AI-first system that empowers human reviewers with better, more actionable information, these companies adopt a human-first approach to review, supplemented by their algorithms. After uploading documents, review teams manually validate all of the information that the AI has identified, fill in blanks where the algorithms missed information, and change fields that their algorithms simply got wrong.
The question for you, then, is whether you’re willing to do all of that manual work to support a less powerful contract management platform. In addition to your own team’s workload, however, there are other key risks you should consider:
Real AI is exponentially faster than humans can ever be when it comes to reviewing thousands of data points. If you need a large group of documents reviewed immediately, can you afford to wait days for an outsourced team to handle it?
Studies have shown that AI is more consistently accurate than humans when performing document review. Not only are you sacrificing time when you entrust your documents to these outsourced teams of reviewers, but you’re also risking a sharp decrease in the accuracy of the review. On top of that, if you don’t have any contact with the human teams doing the work, then you have no way of monitoring or ensuring the quality of their work.
Can you entrust your business’s well-being to an anonymous team of people working with less accurate software?
Another risk is data privacy. If your contract management software provider is outsourcing document review abroad, are they telling you about it? For companies that do business in Europe or California, regulations including the GDPR and CCPA are major concerns. You need to know exactly who is handling all of your customers’ sensitive data, because under those regulations, you’re responsible for the actions that your business partners take with the customer data that you send to them. If you don’t know who is handling your data, then how can you ensure your own company’s compliance with data privacy laws? Even worse, if the vendor isn’t even disclosing to you that they’re transferring your data overseas, then you may find yourself inadvertently in breach of your data privacy obligations.
How can you tell the difference between real AI and the rest?
This is where live testing comes into play.
A true AI system will be able to immediately process and organize a set of documents that it has never seen before. Just upload your contacts to the system, and within a matter of seconds, an AI platform that has been pre-trained on a wide range of real commercial agreements will be able to recognize dozens of standard provisions.
If the vendor insists that they need extra time to prepare the contracts before giving you a demo, that’s a red flag.AI doesn’t take days to read through your contracts. Human reviewers do. You’re the customer, so go ahead and ask: why does the vendor need so much time?
In addition to the AI, a live demo will give you a chance to test other features such as optical character recognition. If the vendor claims that their software can take a flat scan of a paper document and convert it into searchable, editable text, then the system should be able to do that live as well.
A live demo gives you a better sense of what the system’s capabilities really are, and how much manual work you’ll have to do after you upload your documents.
Ready to take the Evisort Challenge? Schedule a demo today!
Find out how
can help your team
Test Evisort on your own contracts to see how you can save time, reduce risk, and accelerate deals.