Checking contracts with artificial intelligence
The startup Legartis offers a software solution that legally analyses contracts in a matter of seconds. This helps companies in reproducing their own guidelines when drafting contracts. This innovative and efficient method was developed at the ZHAW School of Engineering.
Hardly any business is now conducted without a written contract. Depending on the size of the company, legal departments have to review and negotiate dozens of contracts at the same time before approving them for signature. Despite technological progress and efforts to achieve standardisation, contracts today are still checked manually to ensure that all relevant clauses are included and there are no problematic contents that stand in the way of them being signed. For lawyers, this is mostly repetitive and time-consuming work. The startup Legartis is providing a remedy to this problem.
Checked in just a few seconds
Together with the ZHAW, the Zurich-based company has developed software that checks contracts for deviations from the pre-defined corporate guidelines of the customer in question on an automated basis. This means that the contents are automatically identified and classified. The solution from Legartis points out risks and, in particular, identifies and highlights missing and problematic clauses. This is made technically possible by artificial intelligence, specifically by processes from the areas of natural language processing and deep learning, as ZHAW researcher Don Tuggener from the Institute of Applied Information Technology explains: “The software ‘reads’ the contracts, so to speak, and structures them thematically into more than 150 so-called labels. In a matter of seconds, it recognises whether a contract fails to meet a company’s pre-defined guidelines or is incomplete”. Should this be the case, the software points out the erroneous section within the contract.
“In a matter of seconds, the software recognises whether a contract fails to meet a company’s pre-defined guidelines or is incomplete.”
ZHAW researcher Don Tuggener
First application in German
A typical use case for the software is the checking of non-disclosure agreements (NDAs), says Legartis CEO David Alain Bloch, who explains that for his customers the number of NDAs that need to be checked each year can range from a few dozen to several thousand. Until now, software of this kind has chiefly been available for English-language contracts. However, the Legartis software developed at the ZHAW not only understands English, but also German – a unique selling point.
Machine learning
Legartis also wants to stand out from the competition by already applying machine learning, i.e. the training of software with data, in advance and not only on the customer side. “Today, we train the solution for each contract type individually. The AI should also be able to transfer what it has learnt from one contract type to the next”, says Bloch. The finished out-of-box product can be used by customers immediately and independently. “Companies use this solution to allow their specialists to focus on the essentials once more”, adds Bloch. Following the successful market launch at the end of 2019, the young company has since been awarded the title of Early Stage Startup of the Year at the Swiss Fintech Awards.