Our commercial and lawtech practices have collaborated on an enhanced due diligence process that incorporates generative AI. In this article, Martha Crnkovic from the commercial team and Oliver Jeffcott from the lawtech team discuss how the groups worked together and how they see the tech progressing.
 

Q. What prompted you to explore generative AI for due diligence?

Martha: I’ve been a transactional lawyer for over 10 years now. Since AI came on to the scene, I have been interested to see what it could do to remove some of the lower value add work and give us back more time to do what we do best– getting to know our clients, analysing risks, and giving clients the most practical, timely advice possible. In the past, I used more standard technology tools which were less sophisticated than generative AI. When I moved to Macfarlanes, I thought this was the perfect opportunity to test more advanced technologies to see where they could take us.

Oliver: In lawtech, we were looking at opportunities to use generative AI on high impact cross-practice projects. Due diligence was a natural place for us to start, given the volume of routine tasks and number of documents typically under review, a new generative AI-enabled process could lead to significant time savings. It was really helpful to learn that Martha already had experience using technology, including extractive AI, to assist with this. In particular, this meant that Martha was well placed to guide us on where in the process to focus and we were both on the same page in terms of expectations (i.e. AI is not going to get you 100% of the way there).

Q. How did you go about implementing generative AI in your due diligence process?

Martha: We looked at reviewing common or standard commercial contracts, extracting key information, and generating summaries. Initially, we focused on identifying the most critical data points that needed to be extracted and how generative AI could assist in this process. We then worked on optimising the AI to recognise and pull this information accurately. We were also keen to see how well the AI could sift through the data for more “nuanced” information, which would involve comparing terms across the contract.

Oliver: I like to start by understanding what is driving a project. In this case that meant asking questions, for example, does the commercial group want to refine the final output? Does the group want to save time to help improve recovery? Or, does the group want to free up fee earners' time to work on more substantive work? We also wanted to gather more context. For example, how does the task sit within the totality of the workflow? What constitutes a useful output? What expectations exist in terms of time frames for delivery? We used this understanding to train extractive AI models to identify the relevant clauses from the documents. On the generative AI side, we used robotic process automation to enable us to ask a string of prompts across each document. We then called on document automation to input these answers into a spreadsheet or Word document for the commercial group to review. What quickly became apparent, was that the generative AI output was more useful than the extractive AI model for this task. By removing extractive AI from the process, it meant that we could avoid the need to train an algorithm each time we wanted to add a new clause.

Q. What is the key value derived from implementing generative AI to help with due diligence?

Martha: It has been a fantastic learning process. Although I used AI in the past, working with our legal technologists on newer models gave me an insight into the strengths and weaknesses of generative AI that I could not have grasped by reading about it. There has been a lot of knowledge sharing between our practice area and lawtech, which also helps us on the client side. A key learning for me is how important it is to have experts involved in the design and implementation of the AI process. It really is a technical process and having the right people involved is critical.

Oliver: We are now able to create a table summarising relevant clauses across hundreds of documents, which takes around an hour of a lawtech fee earner’s time. Previously this would have taken days or weeks, so technology has freed up large amounts of time that the commercial team can use more effectively and provide more value to their clients. The initial feedback has been very positive, and we are excited about the potential for further enhancements. For the next phase of the project, we are looking at how we can use the summary table to add key information into the final draft diligence report as a first pass for the fee earner to review.  The project has also opened conversations with other practice areas that input into the due diligence process and we are now working with them on similar offerings to support their work.

Q. What comes next?

Martha: Using generative AI for this due diligence process has given us first-hand experience and new-found knowledge of where the advantages and risks are with similar models. We are equipping ourselves with the kind of technical knowledge our clients will expect in deals involving generative AI. It is an exciting time for our team as we are now looking at other uses for generative AI across our work streams, and paving the way for further efficiency and time saving to allow our team to focus on complex work that will add significant value to our clients. 

Oliver: The collaboration between the commercial and lawtech groups has demonstrated the significant potential of generative AI in transforming legal processes. By leveraging advanced AI technologies, we have been able to automate and streamline a complex process, leading to substantial time savings and allowing legal professionals to focus on higher value tasks. There is great potential for even more sophisticated applications. It is an exciting time to be working in this field and we are all energised about what kind of impact the combination of our technical and legal expertise can have on our client services as the technology continues to evolve.