The recent surge in machine learning popularity, largely driven by the success of deep learning since around 2010, has provided much excitement around the possible applications for law firms, explains Rafie, CEO and Co-founder of Genie AI, a machine learning startup.
Understanding artificial intelligence and machine learning
However, there is still confusion in the market about the meaning of buzzwords. Artificial intelligence (AI) is a broad field of study about building machines that are smart compared to humans. In the past, approaches typically involved the use of "expert systems", that is, a set of rules encoded by an expert human. For example: "if green, then apple".
Machine learning, however, involves coding mathematical models such as linear regression, used to show or predict co-relations between variable parameters, in order to learn from past data. Rather than having to manually code "if green, then apple", for example, a machine learning model can ingest image data, then based on millions of parameters that are learned, decide if the image is an apple or for instance a pear.
Measurables / business case
Why are you procuring AI? Many firms procure for the sake of being able to say they are "doing AI". Whilst it is important for marketing to be involved in the latest technology, AI is a tool to enable better business outcomes. It is important therefore to understand the business goals or priorities for the whole firm, then the goals of the team involved, and lastly to measure how the AI technology affects these goals.
Is there a gain in efficiency, a reduction in risk, or an increase in usability? Some gains, like efficiency may be able to be measured quantitatively. Others, like a reduction in pain or a "beautiful, seamless process" may require surveying end users for their opinions. The best firms improve business and client outcomes, rather than ticking the box of having AI.
Fortunately, machine learning is becoming more commoditised and easier to understand. Nonetheless, it is a mathematical discipline rooted in fields such as linear algebra, statistics and optimisation. Not understanding the equations that underpin these models means providers won't be able to build quality machine learning products. It is therefore best to check whether the AI provider has real expertise in the firm - typically this means founders or employees that have degrees in machine learning, and advisors who are professor level in machine learning.
Machine learning, as the name suggests, requires data to learn from. Employing legal AI usually involves a quid pro quo, where machine learning companies exchange expertise and agility for data, adding value for both parties.
Law firms and legal counsels should identify problems and pain points in the company, and look to gather a dataset that can be used by the AI provider to solve those problems. In many cases, the gathering of such data is a problem in itself, and can again be solved by the use of AI, such as automatic categorisation of case law, precedent and documents.
Once data has been acquired, algorithms must be trained. For state of the art machine learning models (often based on sophisticated mathematical modelling known as deep neural networks), this can be a slow and expensive process. It is a good idea to ask the AI provider how they expect to train their algorithms, who will tag or label the data (for some types of machine learning), and how much data is required.
Security is always at the top of the agenda for law firms and it encompasses many different issues. Some issues to consider are: confidentiality, client terms of engagement, data protection, regulation, and cyber security. Law firms consistently find themselves in a catch 22 whereby their clients demand the use of technology and learnings from other matters to drive efficiency, yet also desire the privacy of their own matters.
Ultimately, if clients expect to receive the best possible service from law firms, there needs to be an element of trust and transparency that law firms will use their data responsibly and in a manner that will ultimately benefit them. Furthermore, most terms of engagement permit the sharing of data with 3rd party software providers under certain circumstances, and the success of private cloud enables a private and secure way of hosting data in the cloud, with no connection to the outside internet.
Due to the partnership structure of law firms, it is no secret that procuring technology can be a slow process. Much has been written about agile working, and implementations of agile such as scrum, as well as optimal team sizes and make up to deliver technology projects.
Jeff Bezos' famous "two pizza rule" suggests a team shouldn't be larger than what two pizzas can feed, and agile teams are typically cross-functional, in that they include a variety of different skills in order to deliver projects without requiring much help or sign off from outside the team. This structure would be particularly useful in a law firm when delivering technology projects, since there are typically many different stakeholders, both vertically in the hierarchy and horizontally across departments. A strong cross functional delivery team may have a partner, a small team of fee earners, a knowledge manager and/or professional support lawyer, a product or project manager (although they are very different), an IT manager and a procurement person.
Code needs to live somewhere, and when procuring technology, law firms have the option of on premise, cloud or private cloud. On premise distributions are expensive to set up, require constant maintenance, and are hard to scale. Cloud distributions are scalable, but data may live in databases that also host other types of data. Private cloud marries the best of both of these worlds, by providing the scalability of cloud, and the privacy of on premise. This is because private clouds have no connection to the public internet, and access is limited only to the law firm (and any explicitly accepted parties). Private cloud is fast becoming the industry standard.
Responsibility and liability
Who is responsible for an AI's decision remains a grey area, but it is a question to be aware of. The main factors to consider are, is the AI "assisting" the lawyer, such that the lawyer still makes the final decision, or is the AI replacing a certain process. Is the process business critical, or is it a back office function. If the AI makes a mistake, how much of an issue is that, and are there more or less mistakes on aggregate with the use of the AI compared to humans? What is the cost or gain of the trade off?
Is machine learning really required?
The right question here is, what extra benefit does machine learning provide, that wouldn't be possible either with humans, or with traditional artificial intelligence (a rules-based approach)? Staying true to this question will enable law firms to sift through the hype. When machine learning is truly critical to providing an outcome that couldn't before be achieved, or providing the same outcome significantly faster or more accurately, the benefits can be phenomenal.
Views expressed in our blogs are those of the authors and do not necessarily reflect those of the Law Society.
Our new Lawtech Report highlights key developments and what this means for the work of the profession and the business of law
Sign up for our weekly GDPR and cybersecurity newsletter to keep yourself up to date
Explore our advice and guidance on GDPR compliance
Explore our cybersecurity resources