Artificial Intelligence and Law

Legal hackathons are popping up with such increased frequency, one might think that the legal bar exam would by now include some mandatory tech section. That’s not quite the case yet, but academics and representatives of the legal profession are not tired to emphasize that after retail, insurance and banking, law is the next major sector to get disrupted by tech.

Especially with the emergence of artificial intelligence, the opportunities for true disruption appear to be limitless. Indeed, during the European Policy for Intellectual Property (EPIP) conference 2015 an established academic explained to his younger attendants that he was certain that a substantial number of them would be replaced by AI-driven computer programs. And that prognosis is not an isolated one: at a legal start-up competition for young law students, a speaker expounded that AI would make a number of the students’ future job obsolete and that they therefore would be well advised to learn some coding…

Such optimistic/pessimistic (depending on your point of view) assessments set aside, what kind of applications does AI currently offer to the legal field? How sophisticated in terms of AI are they? How mature are they? How close are these applications to substantially disrupt the legal field?

Taking the classification of Joanna Goodman’s excellent book ‘Robots in Law: How Artificial Intelligence is Transforming Legal Services’ as guidance, the applications of AI in the legal field can be classified as follows:

1) Access to legal services:

This category refers to applications/services that lower the bar for potential clients to find legal services and obtain legal representation. Examples are platforms like Law Gives that enable potential clients to post their legal issues and have lawyers respond to these issues. The maturity of this application in this category varies from experimental to reasonably advanced. The degree of AI sophistication in this area appears to be relatively low. In terms of potential impact, this area has the potential to be transformative, it can empower small law firms and serve segments of society/areas of law that are underserved by the legal market.

2) Legal research:

This category refers to applications, that assists complex research, especially for complex projects with tight deadlines such as M&A or antirust cases. Also, such applications are highly useful for academic research, PhD students or even undergraduate law students. Examples are applications/services such as Judicata, RavelLaw, blueJLegal, Fastcase . Such applications/services can be classified as mature, as they have been used for a number of years and offer differentiated services and pricing. The degree of AI sophistication appears to be low/medium. In terms of potential impact it can help law firms to minimize the size of large back offices/legal researchers/paralegals. However, it seems to rather complement human lawyers rather than to be transformative.

3) E-discovery:

This category helps complex research and discoveries for complex legal cases in M&A, intellectual property, antitrust, financial regulation and general due diligence tasks. Examples are applications/services are Catalyst, DISCO, Relativity, Recommind . The applications/services in this area can be classified as very mature, with many players having been around for many years (partially since the 1990s) and offer highly differentiated services and pricing. The degree of AI sophistication appears to be low/medium. While E-Discovery certainly helps users to reduce time and money spent on research, it might not be classifiable as truly disrupting the legal market or the legal profession.

4) Automation & Scalability:

This category enables to provide legal question in natural language and receive straight forward answers: e.g. ‚Collateral Directive Advisor: Does the EU Collateral Directive apply to this transaction?’ Examples are Beagle AI, Kira, eBrevia , Neota .
Overall, much of the technology appears to be in its infancy and still struggling to truly understand natural language, therefore the maturity of this application/services needs to be classified as relatively low. The degree of AI sophistication appears to be medium-high. In terms of potential impact it can be classified as potentially disruptive – it could enable law firms to scalably service high volume/small margin cases with very few lawyers.

5) Case prediction:

This category helps to analyse/structure highly complex information and helps building models in order to make a qualified prediction on the probable outcome. Exampes are Lex Machina, Loom Analysis, PatentVector.
The applications/services appears to be reasonably mature with some players having been around since the mid 2000s. The degree of AI sophistication appears to be at the higher end of the scale. In terms of potential impact it can help to complement human lawyers, assist making qualified decisions, however it is unlikely to be truly disruptive.

6) Contract analysis:

This category helps to analyse and understand large, complex contracts/service & licensing agreements (for financial services transactions) and prevent costly mistakes.
Examples are RAVN, Kira and Beagle . Overall, the ability of AI to genuinely understand natural language/contract language still appears to be limited. Therefore the maturity needs to be classified as relatively low. While this can help law firms to minimise the size of large back offices/legal researchers/paralegals. However, it seems to complement human lawyers rather than to be transformative/disruptive.

Having provided a brief, most probably imperfect, overview of how AI is applied to the legal profession/market – let us return to the original question: how close are we to a world where indeed the aforementioned speakers would be right when warning young lawyers and students that AI is here to take their jobs?

As shown in the overview, most of the AI applications to the legal field tend to be aimed at complementing rather than replacing the human lawyer. Almost in their entirety they are aimed at taking away/automating ‘boring’/uncreative research tasks that frees up more time for the lawyer to be more creative on the important, strategic tasks of building a case for their clients or helping to decide them what to focus their ‘manual’ research on.

Certainly, some of these applications, such as automation/scalability or contract analysis do have the potential to significantly affect paralegal/entry level junior lawyer positions, but even for those applications to have a significant impact, the technology needs to progress significantly beyond what it offers today.

During the LegalTech Conference Berlin 2017, Dr. Paul von Bünau from idalab provided the following overview/forecast of where the technology is and where/when it is likely to progress to certain stages:

This does not purport to be an accurate forecast – the famous dictum about predictions about the future being tough holds true here as well – but it instead intends give a rough idea how far the technology still needs to progress in order to be able to live up to some of the optimistic claims about AI disrupting the legal sector.

But even if artificial intelligence will reach a stage where it is strong enough to live up to most of the claims, will it drive mass numbers of lawyers out of business? As already hinted, most of the applications are likely to affect ‘uncreative’ researcher/paralegal/junior lawyer jobs. So, on that side yes, the consequences could be disruptive. However, this could have the unforeseen consequence of enabling even more independent individual lawyers/small law firms to enter the market and offer services for which they currently would need to shoulder prohibitively expensive overhead costs. The efficiency gains thus afforded to individual lawyers/small law firms might enable them to litigate cases that otherwise they would have never been commercially viable enough, especially in an age of falling legal aid spending. The consequence therefore might very well be that even more rather than less lawyers being able to make a living. From a social perspective, the efficiency gains might give access to legal services to segments of society so far under-served by the legal market.

Lastly, many of the most optimistic AI / legaltech advocates (curiously enough most of those not being from the AI/computer science field) appear to overlook that the most decisive aspects of law lie in its grey areas. It remains doubtful if AI, however strong, will be able to deal with those grey areas and if any legal system/legal representation would be confident eliminating human judgement from navigating these grey areas.


This article was written by Christian Geib. You can contact him by sending a mail to