The kakemono displaying Alizé’s slogan, photographed at the launch of version 3.0 last February. Photo: Jacques Reiter Photography

The kakemono displaying Alizé’s slogan, photographed at the launch of version 3.0 last February. Photo: Jacques Reiter Photography

Téo Verchère, CEO of Alizé, a Luxembourg-based legal tech company revolutionising access to the law, shares his insights on the future of the profession, having overseen the roll-out of their solution to several hundred organisations in Luxembourg and Belgium.

When Dario Amodei, CEO of Anthropic, claims that AI will replace 50% of white-collar jobs within the next five years, he is perpetuating a narrative which, when applied to the legal profession, is currently nothing more than a pipe dream due to the structural limitations of large language models (LLMs). These limitations have, however, been identified.

Firstly, LLMs do not reason; they predict the most likely next word in a sequence based on the data they have been trained on. This training data is a source of friction because it is outdated, mixes jurisdictions, temporal considerations and degrees of reliability, particularly when it comes to web searches. Players such as Alizé address these issues by providing models with access to structured, reliable and up-to-date datasets, and by making the responses auditable. However, the probabilistic nature of AI generates hallucinations that cannot be completely eliminated.

Secondly, AI operates on the basis of the data it is given access to. However, a crucial part of the work relies on elements that do not exist in any database, such as experience, sector-specific knowledge, customer relations and, above all, critical thinking.

Furthermore, there is no single body of law that could be taught to an AI once and for all. Every country has its own legal texts and traditions, which are constantly evolving. Leaders in legal AI have invested millions in specialised models without convincing results. The idea that simply training a model on the law would be enough to produce a universal expert comes up against this reality.

That said, artificial intelligence is increasingly being deployed, used and valued by professionals. In particular, it enables the drafting of initial documents, large-scale document analysis and searching through vast amounts of documentation. It is on this last aspect that we focus our efforts at Alizé.

Against a backdrop of structural inflation in legal information, AI makes it possible to move beyond the traditional approach of keyword searches, which could sometimes take several hours. It can identify the most relevant decisions in a matter of seconds, compare a set of submissions against a case law database, automate legal monitoring, identify the strategies used by the opposing party based on past litigation, or even estimate the compensation awarded by a judge. It thus provides access to information that was previously difficult to utilise due to a lack of time. Whilst a professional analyses one to two pages per minute, AI can process thousands of pages in a matter of seconds. This ability to process vast amounts of legal information is becoming a decisive advantage for law firms, businesses and institutions in their ability to anticipate risks and make informed decisions.

For more information, visit alize.lu. A full version of this article is also available in French. 

This promotional article was written by Alizé as part of the company's membership with the Paperjam Club. If you would like to become a member of the Club, please contact us at club@paperjam.lu.