“Adopting a hands-off approach to AI in law could have unintended consequences”

While some firms adopt a 'wait and see' mindset, Mishcon de Reya has developed 'deReyAI', its own suite of generative language tools.

It has now been used by two-thirds of lawyers across its workforce since its introduction in October 2023.

We spoke to one of the firm's interactive entertainment lawyers, Harry Clark, on how he uses these tools for daily tasks, the importance of upskilling staff via training, and why pretending AI doesn't exist could have unintended consequences.

Ella Krikler, legal service and product lead, explains how the firm approached integrating GenAI into its daily activities.

Harry Clark, interactive entertainment associate at Mishcon de Reya

I grew up loving the art form of games and jumped at the chance to advise clients operating in the sector. The interactive entertainment industry has always played an important role in the development of emerging technologies such as AI, due to evolving player expectations and the technological demands of creating immersive, real-time interactive experiences.

AI has been around longer than the current 'ChatGPT' zeitgeist would have you believe. My own interest in GenAI, specifically, began a few years ago when I started to experiment with early-stage, novel applications, such as AI-powered text adventure game generators, before then considering how it could be safely and appropriately applied in a work context, as well as in our personal lives.

'deReyAI' is essentially our own internal and private form of ChatGPT, which we developed to ensure the security of our data and to explore early stage use cases. I find it most useful for routine use cases, for example using it to help with administrative issues, or tone - in a drafting sense - and outside the more rigorous confines of a contract. I've found it can help more broadly with the business side of law, such as planning for an event, emailing a colleague or doing blue sky thinking or analysis, acting as a soundboard for ideas.

For me, GenAI is more of a stepping stone to deliver an end product, or otherwise there to give you a slightly different perspective. To capture use cases across the firm, we're also creating a very handy prompt library [that pre-loads instructions to generate specific content or analysis for given use cases] so people have somewhere to share knowledge on what's worked well for them.

Applying GenAI to specific legal questions and tasks is difficult. Its outputs are broadly only as good as the training data, and we're not quite there yet in terms of everyday GenAI models being trained on volumes of high-quality legal data. Lawyers also have multi-stage, specific ways of doing things and, for example, understand the relationship between clauses and provisions in a contract in a highly contextual manner. Off the shelf AI tools don't fully appreciate that yet.

Adopting a hands-off approach to AI in law, or pretending it doesn't exist, could have unintended consequences for both individuals and firms. Without proper training, people might still try to give GenAI a go, but use it incorrectly akin to a search engine, get frustrated and therefore not adopt a change mindset. Or worse, they might use an unsanctioned tool in an unsafe way and compromise the firm's security. We decided to get ahead with company-wide training across the business.

We focused training at a departmental level to address nuances in different teams. There are tasks in real estate that wouldn't apply to a corporate transactional team. So, we've given the teams the tools they need to use the service and craft prompts relevant to them. We also have regular open design workshops so colleagues can share ideas based on what they're working on. Adoption has markedly risen as a result.

Advising clients on AI

To future-proof the industry, universities and other providers must play their part. There is a growing feeling in the profession that university courses and other early-stage educational providers are not modernising fast enough to ensure those qualifying in five years' time will be ready for the profession that awaits them. However, with the right skills and mindset, there is a huge opportunity for lawyers entering the profession to be the next group of changemakers.

Tech spotlight

So how does a firm go about implementing its own GenAI tools?

Meet Ella Krikler, who helped build and roll out deReyAI.

She discusses the ongoing company-wide collaboration at the heart of its AI journey, a security-first mindset, and how her background on the legal team helped inform the specific tools the firm uses.

Ella Krikler

My first introduction to AI - the importance and utility of it - came from an existing close collaborative relationship with the data science team. Our strategy department collaborates with tech and knowledge teams to solve business problems and improve processes for our lawyers. Together, we work to set the tech direction of the firm.

I started my career in the legal team here around five years ago before joining the strategy side of the business. When we're developing new technology tools, it's been an advantage to have hands-on experience of the pressures, expectations and needs of the legal team.

The safety and security of client data is always a paramount concern. We were keen to adopt this new technology early-on to educate our strategic leaders, upskill our colleagues and leverage the existing technical expertise within the firm. However, we didn't begin developing deReyAI until we could do it safely and ensure our data does not leave its secure environment. We stay up-to-date with the regulatory landscape and changes in the technology through collaboration with risk and compliance, knowledge lawyers and senior business leaders. We've also put together a company-wide policy on how lawyers can and should interact with deReyAI.

Start simple, iterate with lawyer feedback

We started by creating a generalist tool, intentionally. Not wanting to restrict ourselves to a user type, we developed our own private chat tool and a document chat which lets lawyers interrogate a set of documents to find information. This allowed us to assess how deReyAI was being used, where our users were getting the most value from its application, and strategically build our roadmap.

Our tools are being used for things such as drafting, redrafting and summarisation. We built a bespoke summarisation tool, which can turn multiple large documents into a concise and meaningful summary, in response to demand from our users. We have a continuous cycle of feedback which informs the direction of our work.

GenAI has helped with productivity. There's a finite amount of time people have to spend on things. Some people want to work from a blank page and then use AI to improve their work, while others start with prompting AI to get a 'first cut' and then add in the layers of expertise. We're changing the way of work to allow you to focus your expertise on the right tasks, making for a better output within the time you have.

"We suggest to our lawyers they interact with our GenAI tools as they would a junior colleague. The better you are at providing instructions, guidance and a framework, the more likely that person is to produce what you're looking for."Ella Krikler, legal service and product lead at Mishcon de Reya

We don't want people to feel intimidated by using AI. We're in a profession where people have a good grasp of language and how to communicate. We suggest to our lawyers they interact with the tools as they would a junior colleague. The better you are at providing instructions, guidance and a framework, the more likely the person is to produce what you're looking for.

In a year, we've reached two thirds of our firm. We want to keep sharing knowledge across the firm so our toolkit is the best it can be. We're incredibly interested in how we can leverage the knowledge we sit on to enhance our AI tools even further.

I want to know more

Have you found this page helpful?

RELATED ARTICLES