Artificial intelligence on the rise

Artificial intelligence will permanently change the legal profession. The more tedious and error-prone routine tasks are performed by machines, the more time humans have to concentrate on complex, challenging issues. The first steps on this path have already been taken.

No company acquisition without due diligence. In painstaking detail, a team of lawyers sifts through the takeover candidate’s contracts to ensure that the buyer does not take any unnecessary risk.

For example, if the contracts contain so-called change-of-control clauses, the contractual partner may typically terminate the contract in the event of a change of ownership. If important customer relationships are lost for this reason, the purchased company is worth significantly less virtually overnight and the purchase price is unreasonably high.

Conscientious computer

What used to take whole teams of lawyers a lot of time, a computer now does in a short time. The computer is not only faster than a human; it is also more conscientious, because a human tires when reviewing large volumes of documents and risks making mistakes. However, the prerequisite is that the computer knows what to look for. This is where artificial intelligence comes into play. The keyword is “self-learning algorithm”.

The change-of-control clause mentioned above is a good example. The clause may have different names or keywords depending on the contract, such as Transfer of Ownership, New Management, or Transfer of Control. Many other formulations are conceivable. In order to work properly, the computer program must also be able to identify formulations that no one thought of when it was programmed.

Self-learning programs achieve this by analyzing texts on several levels. At the first level, the program simply searches for keywords. This search typically yields a large amount of hits. At the second stage, the program searches for recurring structures. This so-called syntactic analysis examines the order in which terms follow one another, whether they are close together or farther apart, and what terms or grammatical structures mediate between them. Many hits of the first level are sorted out here again. At the third stage, semantic analysis, the meaning of the terms used is included. Words with multiple possible meanings are clearly defined.

Following this pattern, the program can also analyze terms that the programmer did not specify. For example, if a clause is called “hostile takeover”, but comparable terms with comparable structures and semantic meanings occur in it as in other change-of-control clauses that the program has already analyzed, it will also flag the “hostile takeover” clause. In the final step, a human can then verify whether the clause is relevant or not.

The amount of data is crucial for success

In addition to the self-learning program, another element is necessary for the system to work successfully: A sufficiently large amount of data on which the program can develop its analysis capabilities. For example, all of a company’s contracts or its entire bookkeeping, or even all of a public authority’s notices.

Beyond the initial example of due diligence, the system offers advantages wherever it is necessary to filter out details from large amounts of data. This could be accounting irregularities, possible compliance violations, or the details of a typical settlement in related litigation. The only prerequisite is that all processes are recorded electronically and thus available for analysis.

Building up the necessary data volumes is a challenge in some cases. Pool solutions will become established here, which aggregate data from different companies or authorities, for example, and thus provide the software with the necessary basic material. The individual participant benefits both through the use of the gradually self-improving software and through the analysis of his data stock.