P2P

68 P E E R T O P E E R : I L T A ' S Q U A R T E R L Y M A G A Z I N E | W I N T E R 2 0 2 3 create text-based content and use deep learning techniques and large data sets to understand, summarize, generate, and predict new content. It is essential to pick the best-fit LLM for a specific use case and challenge your vendor on the matter. Using the best, most cost-effective LLM for each task offers the most flexibility. At Henchman, we have dedicated resources that actively research and assess LLMs to select the best-fitting solution for each use case. This way, your firm does not waste time evaluating different LLMs. Moreover, as we always host the technology on our own servers, we can guarantee maximum security. • Security: Will the vendor use your firm's data to train their model? It's crucial to address potential copyright issues. Also, inquire about security certifications like SOC and ISO and privacy concerns before purchasing. Additionally, having clear policies to handle these issues should they arise is a prudent approach for ensuring a smooth and compliant working relationship with the vendor. • Responsibility: As Maarten Mortier, Head of AI & innovation at Henchman, previously explained, there are different levels of AI maturity, and it's crucial to take it one step at a time (Henchman | Mastering the power and peril of AI in legal tech to reach…). While generative models like GPT and its chat-based or other API-based interfaces hold great potential, they can also be prone to generating inaccurate or hallucinatory information. Careful consideration is necessary to ensure that legal concepts, integrity, and accuracy are maintained when leveraging these tools. Vendors should be acutely aware of this issue and make every effort to mitigate bias in their models, primarily when the output is influenced by it. "Lawyers don't need technical skills. If it appears they need specific training on prompt skills, for example, the AI technology is not meeting its intended purpose and is failing miserably." At Henchman, we firmly believe that the technology lawyers use should be intuitive and able to understand their intentions. Even when it comes to prompting, we believe the vendor's technology should be innovative enough to understand what the user is trying to accomplish. A vendor should invest in prompt engineering and not require end-users to train them in prompting skills. However, incorporating new roles or profiles, such as hiring a Head of AI or establishing a Legal Tech Committee, is a valuable strategy. Within larger firms, the process of continually re-evaluating tools and the adoption process has become a full-time occupation. For example, we have a client in Germany who recently appointed a Head of AI to ensure they make informed choices and effectively challenge vendors. We've also observed a growing trend among law firms in setting up Legal Tech Committees to consistently assess their tools and their adoption, as well as align with vendors' roadmaps and vision. Having these internal skills and structures is vital to being at the forefront of innovation in the legal sector. ILTA F E A T U R E S Siska Lannoo, head of Customer and Partner Success at Henchman, guides clients and partners from small boutique firms to large global law firms through running successful tech implementations. Siska has worked in the software industry for over ten years in various areas, from e-commerce software to sales enablement, which allows her to bring best practices on tech implementations with me to the legal tech space.

• Cover