Artificial Lawyer Spring Review – AI’s Green Shoots

Artificial Lawyer is back from Easter holidays and it seemed a fitting time to do a short review of where we are today in Legal AI Land and where the sector’s green shoots have got to.

Back at Christmas last year the legal AI market was burning red hot with a myriad of companies coming to market and/or finally getting traction with substantial numbers of law firms and corporate clients.

There had also been a sector-wide acceptance (at least among some lawyers) that we had moved beyond the world of talking about AI as something that ‘one day will happen’ to something law firms were adopting and making use of in terms of billable work for clients. In other words, it was all real now.

But, where are we now in the Spring of 2017? As with the previous review, to give you a blow by blow account of every single development and new legal AI and automation company would be a bit long. Instead here are a few themes that seem to be apparent at the moment and that perhaps indicate some of the deeper patterns now forming in the legal AI market.

A (Relatively) Maturing Market

While it may seem to some onlookers that the world of AI-driven document analysis, primarily through the use of natural language processing (NLP), is going bananas (to use a technical phrase), the reality is that the number of doc analysis companies for areas such as due diligence, compliance and lease extraction has not significantly grown since the end of 2016.

We still have the ‘Big Seven’ of AI-driven doc analysis: RAVN, Kira, Luminance, Leverton, Beagle, LawGeex and Seal. There are also eBrevia and Diligen, although we hear less about them over here in London than the above seven. And then we have the AI-driven companies that are still refining their offering, such as Legal Robot, Cognitiv+ and Thought River.

Each company is taking a slightly different path to leveraging legal AI technology. Each is also at a different point in their business plan. Some have moved to having non-founder CEOs already, such as RAVN; others are gaining significant new investment such as LawGeex; and some others are now experimenting with whole new branches of technology, such as Seal, with its venture into the world of blockchain.

But, there has not been a mass of similar companies arriving, despite what some may think. Around a dozen AI-driven doc analysis companies in a $700 billion legal market is by no means an over-saturation of providers, not by a long way. Consider this by way of comparison: there are now around 200 providers of eDiscovery services, and the vast majority are based just in the US legal market.

The real challenge now for law firms and corporates is to figure out which of the above doc analysis AI systems to use, where to use them and what precisely for.

In some cases, firms will be ‘platform agnostic’ and make use of several doc analysis AI systems in parallel, using each company’s tool, or tools, for the task they are best suited to. That is to say, the market is moving beyond seeing ‘AI’ as an amorphous, collective blob that you can just ‘buy some of’, and instead seeing that it is a family of specific tools that exploit a certain type of technology, each with different pros and cons, as well as sometimes quite different use cases. And that’s a good sign. It’s a sign of relative, market maturation.

One other aspect to this maturing is that more law firms are gaining the confidence to develop their own doc analysis systems. Last year we heard about Pinsent Masons building its own DIY legal AI system. This year we have seen Linklaters partner with Eigen to develop unstructured legal data analysis tools, as well as insurance firm, Kennedys, hire a cognitive technology and machine learning expert to take its propriety insurance defence case management system into the realms of legal AI.

All in all this seems to suggest the doc analysis part of the legal AI sector is in a healthy state and that lawyers are increasingly starting to gain a deeper understanding of its multiple uses.

‘Bubble!’ What Bubble?

A new week, a new story about legal AI, and probably another story about an investment in the new wave of legal tech in general. And also probably another story about a law firm, a corporate, or fund helping to start an accelerator or incubator for legal tech companies.

In such an environment it’s understandable that some commentators, whether in the media or in the lawyer fraternity, start to cry: ‘Bubble!’

We have new technology, plus money, plus a market that cannot fully see yet what it’s all about, hence: ‘It must be one of those of 1990s bubble things,’ goes the logic.

But is it a bubble? The short answer is: no.

First, in terms of investment the amount of cash going into legal AI companies is tiny compared to, for example, the billions of dollars invested every year in consumer level web-based applications, or even general business tech applications.

The entire legal AI market would today add up to a fractional rounding adjustment for some of the larger tech companies in the market. In fact, it would be a rounding adjustment on many global law firms’ balance sheets too. That doesn’t make it unimportant, far from it, but it suggests we are nowhere near a ‘bubble’ as such.

Not that many yachts in Legal AI Land

Moreover, none of the legal AI companies are listed publicly, even on the most niche of exchanges. Maybe one day they will be, but we are a long way from that. Taxi drivers are not remortgaging their homes to buy stock in legal AI companies, as happened with the original boom in the 1990s when uninformed investors splashed the cash on new companies they’d received a tip from a friend about.

Some legal AI companies are still primarily self-funded by their founders and live quite humbly on the income they make from their clients. I.e. these are for the most part very stable and well-founded (and sensibly-funded) companies. Where there is external investment it is in the $100,000s to single digit millions of dollars. No legal AI bosses are buying yachts at the moment.

While investment funds are indeed looking at legal tech companies, the total flow of cash into legal AI specifically remains a trickle by wider industry standards.

Is there a lot of talk about legal AI? For sure. Are people writing a lot about it as a subject that is relevant to all lawyers across this $700 billion market? For sure. But is it a bubble? Nope, it is not.


Legal AI has tended to be seen as synonymous with the more famous doc analysis companies noted above. But, doc analysis is only one area of legal AI as a sector.

What we are seeing now is a true diversification of the market. For example, litigation analysis and case prediction systems have truly become an entire sub-sector of legal AI all to itself, with early pioneers from Lex Machina, through Premonition, to Ravel, to CaseText and most recently Metonymy, which has teamed up with litigation funder and advisory firm Fulbrook.

Then we have the legal research systems, where, perhaps surprisingly, ROSS, which has worked closely with IBM Watson, remains with few competitors. Although PenaltyAISearch is doing some great work in the compliance space.

We also have a myriad of legal bots, which it has to be said – and which most of their founders would be happy to agree with – are not using legal AI technology. Some are built around Facebook Messenger, others use more complex and bespoke systems. These are useful examples of semi-automated process applications that can help a person fill out a form, or gain some legal information by going through a Q&A with a set of pre-set questions. But this is not natural language processing at work in most cases, while any machine learning taking place is often also limited.

Legal bots have come in for some mild criticism recently for not being quite as amazing as they first appeared to be. But, this is perhaps down to a misunderstanding of what they do and the technology they use. Hopefully the market will get wise to this and in turn see legal bots as the highly useful, but non-AI applications they are. Also, in time, the word stemming and pre-set response approach to bots may one day be replaced by more advanced NLP technology. We have to accept this is just the beginning and there is a long way to go for legal bots. But, it’s best not to overstate what they can achieve at present.

One legal bot that is perhaps breaking the mould is US-based Loop, which as far as it is understood is the only legal bot to have broken into voice-driven interraction. Because of the use of voice-based language processing this is in the realms of AI, but it remains a rare example in legal bot world.

Then we have the expert systems, which again are not really AI applications. The use of probabilistic inference to prepare the next question in a Q&A expert system is impressive and builds on what many people are doing with bots. But, this isn’t really AI either. But, as with bots, we will see far greater advances in the years ahead.

And we are also seeing advances in other areas such as the great work by companies such as Juro and in automated contract formation, while an effort is now being made to develop a platform to allow users to integrate several AI systems together, (see the work of Larry Bridgesmith and team with Dash).

In short, legal AI is far from just doc analysis. But, equally, not all new legal tech applications that have a degree of automation are in the realms of AI yet, at least not from Artificial Lawyer’s humble perspective. But, given that legal AI as a sector only really got going commercially in the last couple of years there is everything to play for and what may appear to be relatively simple apps today, may soon enough become far more advanced. If there is one thing that technology has taught us, it is that it rarely stands still.

And the overall message remains: the legal AI market is diversifying and becoming far more variegated than before in terms of use case. All of which is good news.


What we have as of Spring 2017 is a healthy, growing and increasingly diverse sector of ‘new wave’ legal technology, with AI at the vanguard and a mix of other technologies, ranging from word-stemming bots to blockchain, taking up the wings of this onward charge.

There doesn’t seem to be any bubble. What Artificial Lawyer sees is a lot of very creative people developing great ideas designed to make the legal sector a better place for lawyers to work in and for clients to get a better delivery of services from, and in some cases an improved access to justice. And that sounds good to me.