Embedded Law

by Charles Kerrigan*

Abstract

• This article is a thought experiment on the future of legal services.
• Technology is changing, and changing us.
• It will have implications for the work of lawyers.
• Our natural assumption is that if change is incremental, it isn’t radical.

Introduction

We always overestimate the change that will occur in the next two years and underestimate the change that will occur in the next ten. (Usually attributed to Bill Gates)

The latest AI systems are a new set of technologies capable of profoundly changing the world in which lawyers and our clients operate. 

A recent presentation by Gartner, a consulting firm, predicts that by 2027 94% of law firms will have adopted generative AI. 

Perhaps we will overestimate the short term change that will occur as a result of adoption of generative and other AI technologies in legal practice. 

But if we are underestimating the nature and quality of the change occurring over ten years, in what ways might we be doing so? 

The answer is likely to involve us failing to update our frames of reference for how the technologies will be operating.  Let’s try to do that here.

This is an optimistic view.  For those who read these things, a Techno-Optimistic view… The Techno-Optimist Manifesto | Andreessen Horowitz (a16z.com)

Capabilities

We should start by setting out some capabilities of the latest technologies.

1. First, the computers can read and write now.  In my view, this is an existential change.  Lawyers describe their jobs in many ways, but we’ve all made a living from reading and writing.
2. Second, the computers can do things for themselves.  Agentic AI systems operate independently, making decisions and taking actions without direct human intervention.  They are proactive in working toward goals, and learn from their environment and experiences, improving their performance over time.  Already they show human-like reasoning capable of managing multi-step processes.  
3. Third, the computers can be dispute avoiders.  Google’s “Habermas machine” is an AI model designed to minimise and resolve disagreements, to help groups find common ground on divisive issues.  AI generates consensus by averaging individual positions which avoids direct interpersonal and emotional conflict.  Participants in the tests preferred this approach to human-generated resolution of issues.
4. Fourth, the computers are multi-modal.  They can read and write, but they can also work with voice and images (like people do), and raw data (like people definitely don’t).

And, we can connect that to a big trend in business: how technology helps business find customers at their point of need. 

The finance industry is being transformed by embedded finance and embedded finance is transforming other industries. 

Embedded finance is the term used for the placing of financial products within non-financial (online) environments. E.g. if I buy a pair of trousers, I can pay for them using credit in the merchant’s store – i.e. without talking to my bank (or opening its app).  I can insure my new trousers against the risk of losing or damaging them, again in the merchant store – i.e. without talking to my insurance company or opening its app.  In the context of e-commerce, it is an improvement in customer service.  But really, it says to people that products (and then, in my view, services) will come to them, not the other way round.  “Customers benefit from contextual, seamless experiences,” according to Adam Davies of Bain, a consulting firm (quoted in a Payments Association article in 2023).

To help us consider what this looks like put all together in the context of legal services, we can ask some questions of the more traditional type.

What do clients want (why do people call lawyers)? 

When asked why their clients engage them, lawyers generally refer to clients needing some or all of these things:

• Legal advice - to know that they are not and will not inadvertently breach a law or regulation.
• Market practice – to know that their contracts, policies, or approaches to regulation are not inadvertently different from their peers.
• Judgement – to obtain the best answer when there is no right answer; although some of this is reputational rather than factual: if there is no right answer, a lawyer’s answer can only be partial.
• Accountability, and professional indemnity insurance cover – to have recourse when things go wrong.
• Trust – a sense that their issues are being handled within a positive relationship characterised by consistency, doing what you say you will do, and showing expertise and good judgment – people trust those who have good answers and solve difficult problems.

Reflecting these back through the lens of how improved AI systems, and our ability to best use those systems, will change over the next ten years:

• Legal advice – since AI systems can review and understand text, they are excellent candidates to identify and analyse laws and regulations – legislators and regulators are not hiding the rules, they want them to be found(!)…and it is likely that regulators will adopt supervisory technology that uses AI systems to do real time testing within our ten-year horizon (see ‘The Cambridge Regulatory Genome Project: a must in a world of regulation technologies’ (2023) 9 JIBFL 629). 
• Market practice – this is an information problem: market practice is shorthand for the sum of data making up the terms of transactions in a market; no human can compete with a machine on an information problem.
• Judgement – judgement informed by data will be viewed as more valuable as AI systems become transparent and verifiable.
• Accountability, and professional indemnity insurance cover – remember the point above about embedded finance and insurance.
• Trust - we don’t trust AI responses now but (please excuse the clichés) (a) ten years is a long time, (b) the AI we have now is the worst AI we will ever deal with, and (c) the rate of improvement between Chat GPT 3.5 and Chat GPT 4o (2 years) suggests we can be optimistic

The last point is the most important and the most interesting.  Lawyers, who are people, assume they will be dealing with clients who are people.  But, in our ten year horizon, there is a lot of automation coming…

I’m not saying that there is no room for people, far from it.  But it will be the people who have good answers and solve difficult problems using the best available tools who are in demand.

Search

Search is an example of the potential scale of disruption coming from AI to information services industries. 

Search engines have to engage with AI, in particular generative AI, because it is a fundamental disruptor for them.  The business of selling ads next to links relies on search results being links.  If the links disappear, the ads won’t be seen. 

The future of search is:

• Conversational: don’t type keywords, ask questions and get coherent, context-aware answers.
• Personal: results that reflect individual preferences, behaviour, and past interactions: i.e. more useful and user-friendly
• Deep: systems that understand complex queries, weigh different perspectives, and provide contextual responses (instead of keyword matching)
• Integrated: with voice assistants and augmented reality: i.e. a seamless experience of information delivered through different interfaces and devices – remember the Adam Davies quote above
• Trusted: AI systems will ultimately be judged on their accuracy and trustworthiness
• Proactive: anticipating user needs based on the user’s current context and activities.

Now the last two points are the most important and interesting.  Push and pull.  Law firms currently largely rely on clients contacting them for advice.  Clients pull the advice from firms.  The better alternative for clients is for advice to be pushed to them, from a trusted source. 

What do clients really want? 

This is an old one, told to me many years ago when I was starting out as a lawyer.  (The name of the well-known magic circle partner I worked for who said it is available on written request.)

Clients want the four As…Availability, Affability, Affordability, Ability……in that order.”

Next generation AI will be…

• Proactive
• Friendly
• Priced inside other services you are buying
• Subtle and correct

The future of law looks like the future of search.  If search is the future, who can build it?  Big law or big tech…?

*Charles Kerrigan is a partner at CMS in London. He is the author of The Financing of Intangible Assets, TMT Finance and Emerging Technologies (2019) and author and editor of Artificial Intelligence Law and Regulation (2022). Email: charles.kerrigan@cms-cmno.com.