Artificial intelligence searches contracts

AI for legal departments 

No more sifting through files

Artificial intelligence not only helps lawyers to research precedents, it also provides support to other company departments. Soon it will no longer be necessary for employees to sift through tens of thousands of contracts to find information – they will simply be able to display the documents that they want on the dashboard in a matter of seconds. If they want to, they can then analyse them in depth with their technical or legal expertise. Sounds like something from the future? Not at all – much of this is already a reality.

Text: Ladina Camenisch, Images: ©Keystone, 14 December 2017

Patricia recently passed her law exams and is now working at a well-known law firm. However, the work isn’t quite what she was expecting. Instead of going to court, she spends most of her time working through old documents, searching for precedents or comparing old contracts. Patricia is not alone: The law firm employs a number of junior staff whose principal job is sifting through files and contracts. While the work is important, progress is slow and taxing.

Now the first law firms in the USA are using algorithms to do Patricia’s research work. Because artificial intelligence (AI) can capture, analyse and collate vastly more data than a human being. The first reports are positive: AI is very thorough, the law firm saves a lot of time, and the next generation of lawyers can take on other, more important tasks.

Analysing 70,000 contracts

Although it’s not a law firm, Swisscom is also confronted with similar problems. Over the years, countless documents have accumulated throughout the Group. And new companies continue to be added, making things ever more complicated. While the signed documents have all been archived and are available as PDFs, how do you locate a specific piece of information? The answer to this question is complex, because different people within the company require a huge variety of information, and often in some detail. Does the contract really state what the customer says it does? Were special GTCs agreed? When does the contract expire and need to be renewed? Does it still fulfil the regulatory requirements? Are there dependencies on other contracts?

The needle in the haystack

At present, Swisscom possesses more than 70,000 contract documents, and around 6,000 new documents are added every year. René Gamma, Project Manager at Swisscom Group Steering, describes the situation: “We have to analyse these documents in a structural way. Only when we know the risk and search patterns can we initiate the necessary measures.”

René Gamma, Projektleiter at Swisscom Group Steering

Another problem is that signed documents are often stored on different systems, sometimes as scanned images, sometimes without meta data, and often in different versions. So the search for binding information is like trying to find a needle in a haystack. In the past, employees took many days to locate the right information within this huge volume of data. Fortunately, artificial intelligence and machine learning now provide valuable help as they autonomously recognise patterns and also learn as the search yields results.

More than just a text search

Andreea Hossmann, Head of Data Science at Swisscom Enterprise Customers, helped program the algorithms. “We first talked to our colleagues who are affected by this in order to understand the existing processes and identify the difficulties. Then we asked the question: How can technology help you in an intelligent way?”

Andreea Hossmann, Head of Data Science at Swisscom Enterprise Customers

The answers enabled them to define a number of goals for the feasibility study:

The solution should be able to …

  • digitalise all historical documents and display them on a dashboard with just a single click.
  • index documents, filter information based on available meta data, and search it for keywords.
  • find related documents automatically and identify the language and specific contract sections.
  • set up categories and related hierarchies automatically.
  • classify documents automatically according to topic.
  • use additional intelligent functions based on AI.
  • ensure the confidentiality of the data.

The learning part of AI is based on what are known as document embeddings. This method is fairly generic, but the software was designed specifically for contract analysis. “It would also be possible to use the method for technical documentation or something similar,” Andreea Hossmann explains. “However, we have to configure the final software product specifically for each application scenario based on a Java library for deep learning.”

The results are convincing

The feasibility study was a success – the search for specific documents and elements is extremely efficient. “Automatic Contract Analysis now enables us to perform comprehensive contract analyses in a reasonable amount of time. This was previously almost unimaginable or only possible with a huge amount of work,” René Gamma explains. “The search is very intuitive and we were already getting good results during the test phase.” Another advantage is that the longer Automatic Contract Analysis is used, the better the system becomes, because it never stops learning.

Automatic Contract Analysis in action at Swisscom

Customers have already expressed an interest in the solution. On the one hand, they want to have control over the existing documents, and on the other they want to work faster, more precisely and more efficiently than their competitors. Additionally, the increasingly strict regulatory requirements mean that all contracts must be continually monitored, and this is hardly possible any more with realistic personnel resources.

And what about Patricia? She’s looking forward to AI becoming a fixture in Swiss law firms. Then she’ll be able to put slow, routine work aside and dedicate her energies to the things that require a lawyer rather than a machine.

One-Stop-Shop for AI

The Swisscom competence centre for applied artificial intelligence offers companies everything they need for the quick, successful implementation of projects in all areas of artificial intelligence, from consulting to the right technology to integration.

> To the offer


Subscribe to the newsletter now and stay informed about trends, industry news and benchmarks.

More on the topic