In the legal world, knowledge is fundamental for the configuration of an AI model. With the emergence of systems like ChatGPT, the legal sector is one of the most impacted. This is well understood at Lefebvre, a portal dedicated to all types of legal information. Their database accumulates legislation, jurisprudence, and even current news about the sector.
After all, large language models (LLMs) deploy linguistic capabilities. It’s important to remember that the raw material a lawyer works with is language. We discussed all this with José Ángel Sandín, CEO of Lefebvre Spain, who has experienced the transformation driven by AI over the past few years from a leadership perspective.
Is the impact of generative AI due to the improvement in the quality of work or the time savings, which allow for resource liberation?
What happens is that a lawyer has a process of discovering what is correct, finding the correct information to draft opinions, documents, claims, or contracts, and to advise their clients. Then there is the actual task of constructing that information to present.
AI has greatly accelerated the knowledge phase, the understanding of that information. As long as you work with systems like GenIA-L (Lefebvre's generative AI system), where you create a perimeter of information for the system to query. We have a very powerful search system that returns very relevant results for the generative AI to work with. For example, in resolving a legal issue, a lawyer might spend an entire day just researching, reading different documents to extract the interesting pieces. We can save them hours and hours of work in seconds.
Does GenIA-L do this thanks to Lefebvre's database?
Yes, they are RAG (Retrieval Augmented Generation) systems, which are increasingly recognized in the tech world. It means that ChatGPT, for instance, doesn’t operate across the entire internet but within this perimeter of information. This is an RAG system. And it ensures that the responses do not hallucinate.
Can you give us an overview of how GenIA-L was developed?
When ChatGPT appeared in late November 2022, it took us only four months to have a solution ready, which is the same one we have now. It has improved in quality and skills, but in essence, it’s the same. We could do it because we already had the knowledge base and technology base ready to incorporate this technology. We had structured databases and high-value-added information, like mementos (legal analysis guides), over 100 titles written by human intelligence, extracting the important parts of the regulations.
What’s the future path for this tool?
This is the first step, but we are already taking other steps, such as document review. We already have a solution there. We haven’t launched it yet, but it basically involves the client uploading their information and us interacting with those documents to help validate, verify, and ultimately help create their case. Not just starting from the information we provide but from their own knowledge. And this is tremendously useful. Because a lawyer or a firm, when they approach a case, usually starts with a document that was already created, where the know-how of that firm is stored. Over years and years of legal practice, many lawyers working together, they form a documentary base, and in those documents, knowledge is stored.
It’s something similar to what PwC would do as a client under the agreement you have with them and Harvey. What does it entail?
Here we provide the knowledge and search technology. Harvey and PwC, without this search technology, couldn’t validate any document or information, couldn’t say if it’s correct or not, any information or question asked. We provide that access to the correct legal information.
How is the value of Lefebvre’s database planned to be managed in the future?
This comes naturally because we have a very extensive client base. We are leaders in the Spanish legal information market, and this gives us a huge possibility to spread our tools, because there are clients who already use our information. From a go-to-market perspective, the step is easy for us. We just have to tell them that an extension of the product they already have will make their life much easier.
It’s a time-saving that translates into cost savings…
A client, instead of handling, for example, 20 or 30 cases a year, can handle 60 cases easily, because their understanding of the case and ability to provide solutions multiplies. This is in a small firm. In larger firms, they have an army of documentalists, juniors who analyze information, and these juniors can now move on to more specialized tasks.
It will also be a revolution impacting how young professionals enter these types of organizations. They will have a much more exciting progression within these consultancies.
One of the concerns is how these juniors will be trained now that they won’t do that fieldwork?
When I did my thesis, there was no Internet. I had to go to universities, to the Fulbright Foundation, to the Sorbonne University. I went to the library, explored documentation centers to get documents that would feed the bibliographic base on which my thesis would rely. It was a whole job. I remember spending about six months doing it. Now we do it in minutes. The intellectual and physical time I invested in that, a student now invests in learning faster.
What will happen is an acceleration of learning. Obviously, through ease of access. It’s like the calculator. Until the calculator existed, things were slower because you had to do the operation. But once the calculator arrives, the operation becomes administrative, and you focus on what adds more value. The same thing will happen here. There’s a period of adaptation, but given how quickly this technology has been adopted, I anticipate that this adaptation period will also go very quickl