Legal tech and AI: best practices from Schoenherr's banking & finance team

30 November 2022. Overnight, AI becomes a commodity accessible to everyone. Within a short period of time 152 million users had interacted with OpenAI's ChatGPT. By April 2023 this number had increased to 1.8 billion, making ChatGPT by far the fastest growing tech product in the last 20 years. When thinking about legal tech one tends to forget that AI has been part of document review and e-discovery products for more than four years now. Why did it not make an impact sooner and why all the hype around AI only now?

AI experiment

As part of our AI experiment in roadmap24, we have curated a few prompts and asked AI about this article. Take a look and find out what ChatGPT responded*:

Previous generations of legal tech solutions were based on AI using machine learning (ML) technology and large language models (LLMs), which are vastly outperformed by the current capabilities. Moreover, the expectations for this technology were extremely high and often unrealistic. Users expected these solutions to replace entire review processes and thus the gap between expectations and capabilities was simply too high. Add the challenges of adopting new technologies and a reluctance to change established workflows and one can see why these solutions had little chance of making a major impact until now.

AI's language-centric capabilities have substantially improved, with actions like classification, extraction, summarisation, simplification and even drafting approaching, if not exceeding, human performance in some domains. This is why we believe AI will have a substantial impact on our profession and will provide an opportunity to further develop the value we provide and the way we deliver our services to our clients.

How are we embracing these trends at Schoenherr? The answer is twofold. We have been using the momentum fuelled by AI to further consolidate the areas which we consider to be the "classics of legal tech" where technology can provide additional value to our workflows and services. Additionally, we have been testing and experimenting with AI while also focusing on education and making sure colleagues and clients understand the new capabilities of this technology without losing sight of its challenges and limitations. We would therefore like to take you through a short journey across some of our legal tech use-cases at Schoenherr.

Along with the usual collaboration tools that allow for faster coordination and co-authoring capabilities when drafting documents across multiple internal stakeholders, we use a state-of-the-art transaction management solution. This platform can facilitate the collection of documents, comments and approvals from various parties, instead of circulating documents by e-mail and painstakingly compiling comments. It can be used to produce issues lists, document lists, KYC lists, etc., while templates of such lists can be easily imported and adapted. The platform gives real-time access to all parties with the possibility to grant different permissions to make changes and approvals. It also provides an overview of progress and outstanding items as well as an audit trail of all changes. Signing can be facilitated through the platform with the integrated electronic signature function or by uploading wet ink signatures. Transaction bibles can be generated with one click.

This solution is constantly being developed further with new integrations and features like automatic document version comparison or more complex notification and signing features. While Schoenherr is using this platform across practice groups with multiple clients, we would like to highlight one of our first use cases: how our banking, finance and capital markets team is successfully applying this solution to manage complex financing transactions.

In a financing transaction, the borrowers and the guarantors are required under a facilities agreement to provide numerous documents to the financing parties prior to and/or following the flow of funds (so called "conditions precedent" and "conditions subsequent"). These documents include constitutional documents, corporate resolutions, directors' certificates, security agreements, including any evidence of the perfection of the security, legal opinions, financial statements, fee letters and others. Once the full set of required documents is agreed upon, the list is imported to the transaction management solution with responsibility, tailored access, review and approval rights assigned for each document. The parties can then directly in the platform upload documents, comment on and make any other changes, which are all tracked. The agent, acting on behalf of the financing parties, reviews and approves the uploaded documents. This transaction management solution significantly streamlines the document collection and approval process, eliminating the need to circulate documents to all parties by e-mail, compile comments, manually keep an overview of the process and request approval by the agent. After approval, all documents can be easily downloaded in a zip file as a closing bible.

Document and contract automation is one of the initiatives that we are currently further developing within Schoenherr. We see a huge potential in automating our know-how to better and faster create first drafts of documents, in order to focus on the more complex work of tailoring these drafts to our client's needs. Additionally, we provide automated templates to our clients, while also advising and supporting in-house departments on automation, which our clients in turn can utilise to provide self-service business capabilities. Although the process of the initial automation of the standard set of documents can be quite complex, the speed and simplification of the subsequent automated output greatly improves the workflows. The initial automation requires proficient knowledge of the sophisticated document structure and a mix of the suitable automation solution and the involvement of the right stakeholders.

A perfect example of the current hype and of AI capabilities not being fully understood is related to automation. We are frequently asked by clients and peers if automation is now dead because of generative AI. The answer is an emphatic "no". Why would we want to rely on a partly random draft that is slightly different with every prompt instead of capitalising on our automated best-practice know-how and templates reflecting years of experience and fine-tuning? While AI will certainly boost the automation process, it will not replace the concept altogether. To get a better understanding of the potential impact of automation, we again refer to a specific banking and finance use-case around the creation of corporate documents and security agreements for a financing transaction.

A standard set of corporate resolutions, directors' certificates and security agreements is required for the borrowers and the guarantors in a financing transaction. Once this set of documents has been initially automated, it can be used in any transaction. In the first step a questionnaire with legal and commercial choices as well as a description of the parties is filled out and the scope of the documents is determined. The programme then generates tailored documents based on the questionnaire answers. Any other necessary changes can be made manually. The answers can also be revised and the changes are implemented straight away in the documents. This process helps save considerable time and resources, especially given the number of parties involved and the usually tight deadlines. More complex standard financing documents, such as Loan Market Association’s facilities agreement and intercreditor agreement can be also automated.

We are currently testing and using a wide range of tools, some powered by AI, that enhance how we access our know-how and that support us during the review and drafting processes. For example, one solution leverages AI and the use of powerful semantic search to allow finding and comparing definitions and clauses in various documents contained in our internal document management system. We are thus able to find the proverbial needle in the haystack by pinpointing a specific clause from previous cases. Features like filters, tags and smart automatic classification allow us to work with the wealth of internal know-how in a structured and time-efficient manner. This software can also create document outlines and answer certain questions about the information contained therein. These AI-powered know-how tools are rapidly developing and we are currently testing new solutions and capabilities so we can stay ahead of the curve and utilise the best technology when it comes to improving internal workflows and service delivery for our clients.

Finally, we are using a wide array of utilities and tools that support our drafting and review processes and can assist with the proofreading of final drafts, like spellchecks, defined terms, brackets, cross-references, incomplete sentence structures, etc. We use smart tools for project management, mapping out processes and diagramming, coupled with by now standard solutions like electronic signature, VDR and document sharing.

All the tools and solutions discussed above help us reduce errors and save time and resources while providing services to our clients. Even though AI has the potential to enhance these solutions and the respective workflows, we firmly believe that our know-how and our personal interactions with our clients form the true backbone of our services.

authors: Ursula Rath, Viktoria Stark, Andrei Salajan