News You Can Use

Research by Anthropic into how Claude 'thinks' when solving problems or carrying out instructions. Some really interesting and unexpected findings, including that the 'thoughts' shown in chain of thought explanations may be faked reasoning rather than the steps actually carried out by the LLM.

Understanding how these models truly think and work will be crucial for building trust and reliance on their outputs longer term. The issues with LLMs telling humans what they want to hear and 'faking' chain of thought means that even if the answer is correct, not showing the working may cause issues, especially in legal. This is specifically around Claude, and we are not sure if other LLMs will reflect the same findings, but it shows there is more than just a 'next word prediction' process running.

The Lawyer spent time with 30 law firms discussing approaches to innovation, including Kerry and Elliot. The full report is due out for the next NYCU but high level findings showed that firms are investing more than ever in R&D efforts and new technology. There is a fear of being left behind but also a fear of acting too soon and taking too much risk, elements of 'innovation fatigue' are still present in the industry but some future proofing around business models is starting to happen.

The broad findings from this report seem to indicate that large law firms are not moving that quickly, and at AG we are probably quite a head of many in the terms of the conversations we are having with both our clients and partners. Some of the wider business model change is coming from the likes of A&O, Macfarlanes, Simmons in selling AI enabled services - but these do not seem to be driving large amounts of revenue (or profit). The wider shift towards an R&D mindset is showing that many other firms will be building the capabilities to catch us up, speeding up the wider change in the industry hopefully.

A report from Juro based on conversations with over 150+ lawyers. A staggering 99% of in-house lawyers believe AI will change their job within the year. 90% of those surveyed actually used an AI tool daily or weekly. Lawyers have a clear mandate from the business to use AI more in their work. However, there is also a lack of understanding around some key concepts and tools. Perhaps the most important stat is that 45% of in-house lawyers think that their firms offer poor or terrible value for money...

This echoes what we are seeing from our clients, mounting pressure to use AI alongside a lack of understanding. Ideally we can push on this to support the roll out of our AI Workshops and training, but we are also seeing this in pressure on our fees and how we carry out work for clients. Interesting redlining as a feature is something lawyers seem to be unable to do but want to do more of - highlighting the importance of properly building processes around playbooks and risk flagging. The attitudes towards value from law firms at the moment show that there is a huge amount of opportunity to deliver more for our clients, but balanced against the difficulty of unreasonable demands and misunderstanding of tech.