The RAG Report: Can Large Language Models be good enough for legal due diligence?
There is a huge opportunity for the application of Generative AI to legal work. At Addleshaw Goddard we have been embracing this new technology over the past two years, building out our understanding and AI capabilities. Following months of research into the validity of Large Language Models and a Retrieval Augmented Generation approach to legal, we are sharing our findings to provide insight into the realities of using Generative AI in law.
This comprehensive research report, which we believe is the first of its kind from a law firm, sets out the work carried out by our Innovation Group to develop and test a robust method of using Large Language Models to review documents in the context of an M&A transaction legal due diligence project.
Our research includes learnings from developing AGPT, Addleshaw Goddard's internal LLM based solution for chat and document review, the steps being taken to develop an M&A transaction specific platform and the optimisation process taken to get our RAG solution from a performance of 74% accuracy to upwards of 95%.
Some of the key findings
The outcomes of this research and related work show that you can make a start with the knowledge you have today. With a mixture of expertise in the technical and legal aspects, and without significant expense, we have been able to develop a sophisticated platform without reliance on largescale data gathering or structuring up front and without training specific models.
Key contacts
Our expertise
Our Innovation and Legal Technology Team
Our passion in this area is driven by our clients and their business challenges. We develop, test, invest in and embrace new technology on a continuous basis in order to enable the smart delivery of legal services and creation of clever solutions to clients, both faster and more cost-effectively.
Find out more