DIY Legal AI – The Pinsent Masons Way

Earlier this week UK-based law firm Pinsent Masons announced it was using AI technology to provide clients with a Brexit risk analysis tool. The story duly received some moderate attention, but didn’t exactly set the world alight. This was perhaps because RAVN along with Dentons and its technology platform Nextlaw Labs had teamed up together in July this year to do something not that different already.

Using an AI cognitive engine to examine a client’s contracts and then extract key clauses to check against any potential legal or compliance risk caused by Brexit is an interesting use of Natural Language Processing and machine learning. But, that’s not what is most interesting here.

The really interesting bit is that the AI system used at Pinsent Masons, known as TermFrame, is home grown. It’s a piece of Do-It-Yourself legal AI, so to speak.

The firm states that: ‘Pinsent Masons uses its own AI platform. This AI technology has been developed over a number of years and successfully deployed.’

David Halliwell, Director of Knowledge and Innovation Delivery, explained the firm’s strategy: ‘A number of law firms are partnering with third party organisations to consider how to take off-the-shelf AI technology and apply it to legal challenges faced by law firms.’

screen-shot-2016-11-10-at-20-37-31We have taken a different approach, beginning with the challenges clients need to address and developing an AI solution to suit through our own R&D team. The technology has been successfully applied across a number of transactions,’ Halliwell concludes.

When one considers that many of the largest law firms making use of AI-driven cognitive engines are using systems developed by Kira, RAVN, LEVERTON, Luminance, Diligen and several others, then this is quite a departure.

It also raises some interesting questions, such as: why bother to go DIY when you can just licence the software and pay £x per document analysed to a legal AI company? Why bother to spend so much time and money, especially on technical staff with the right coding skills and experience, to build this kind of system? Why make one’s life difficult when you can, as Halliwell notes, bring in third party software?

Here are a few possible answers:

  • No more paying for those AI software licences and document review fees.
  • Once the sunk costs of developing the AI have been eaten up by client fees, and after staff costs and some other operational running costs, the AI could be turning a nice profit.
  • Because the firm owns the software the tech team will learn a lot and keep learning the more they use it, not just how to use it, but really get ‘under the bonnet’. That could be very useful in developing new kinds of AI applications and AI-driven products.
  • The system, if one day the partners wanted to, could be sold off/spun out into a separate business that would provide income to the partners.
  • Building your own AI empathically sends a very positive message to the client base about your commitment to technology and efficiency.

There are undoubtedly other positive reasons for taking up the DIY AI strategy. Yet, on the other hand, there are also a few considerations for firms in general that may choose to go down this route:

  • Does another AI cognitive engine for document review need to be built when there are several already in existence that focus on the legal market, with more appearing every few months? Why not focus on building a new type of AI capability that no one else has designed yet, but for which there is an untapped client demand?
  • Even if you have trialled other AI systems beforehand to get an idea of how their systems work, their software will keep developing and advancing beyond where you have got to. For example, several AI companies bring out a new iteration of their software every year. Will a law firm’s DIY set-up be able to keep up with tech companies that are focused solely on AI development?
  • Given that AI, rather like the law, is a knowledge intensive business, what happens if your machine learning experts all take a walk to somewhere else? Does the ability to develop the DIY AI come to a grinding halt? Or do you just keep paying them more and more to keep them in the firm? (Although a recruitment battle for legal AI talent will probably develop in any case, given the sector’s growth and affect most AI companies as well. So perhaps this is a moot point.)
  • Does the law firm have sufficient volume of activity to merit the investment and ongoing staff costs of having one’s own personal brand of AI? In a large transactional firm or a firm where there was plenty of compliance/risk work to be done, then the answer would be yes. But, clearly, there is a scale issue for going DIY on advanced legal tech. [N.B. Pinsents’ annual revenue 15/16 is publicly quoted as £382m ($479m), so presumably the answer is yes in this case.]
  • Will the AI companies want to work in the future with a law firm that has ‘gone DIY’? After all, why play nice with what is now a potential competitor? On this point there is no evidence one way or the other, but it’s an interesting one to consider.
  • And, perhaps the final point is: will what you make in terms of end product be as good as what is on the market? Presumably if you have tested the other systems and can then test your own, then one can get some general indication of comparative performance. [Although how one accurately compares AI systems is a whole other debate…..]

But, overall, the ambition to build your own AI has to be admired. It is clearly innovative and also shows a type of entrepreneurial spirit that the law is not that well known for, or at least was not known for in the past.

And we should also take note of the work of a smaller UK law firm, Hodge Jones & Allen, which is developing its own case prediction system for personal injury matters. Meanwhile, the work of Dentons and Nextlaw Labs in investing in, fostering and guiding to fruition new legal tech is another example of law firms not waiting to be handed new technology on a plate, but rather going out and trying to help it to grow.

To conclude, this is an important step in the development of the legal AI market and Artificial Lawyer will be watching it with great interest as the story continues to unfold.