The future is here - it's just not very evenly distributed
With this post, our Chief Scientific Officer and legal engineer, Dr Ben Gardner, begins a series of articles that are focused on education and a legal engineering capability model. This first post sets the scene by speculating what the future of some aspect of the legal landscape could look like, based on what is already technically possible. Future posts in this series will introduce Wavelength's new 'capability model' and explore its four main components - text analytics & extraction, document automation, expert systems and data visualisation.
As a Legal Engineer, I am frequently asked to predict what the impact of AI will be on the future of law. I have many problems with this question, not least that making predictions of the future is a fool's errand, but also the implication that the future is influenced by a single technology, Artificial Intelligence (AI). It is my opinion that the future is never defined by a particular technology but instead how a number of technologies and data combine to influence society and business. Furthermore, there is inevitably a lag between the emergence of a technology and when the opportunities it enables can be realised. For more on this, see Gartner's hypecycle.
In the words of William Gibson, "The future is already here — it's just not very evenly distributed" and so, in this post I am going to respond to the question about what the future could look like, but I will limit this speculation to one based on existing technologies.
A possible future
The diagram below illustrates how various technologies might come together in the future to transform financial markets.
In this future world, contracts are drafted as structured documents that can be read by humans and computers. Not only are these contracts machine readable but also machine executable (i.e. smart contracts), and include within them explicit references to authoritative data sources that provide real time information that trigger provisions within the contract. These smart contracts could be stored in a shared repository, which could be accessed by all parties and the regulator. Just as a smart contract can be created, it is also possible to create a 'smart regulation', i.e. a regulation that is both readable and executable by humans and machines. Smart regulations would allow the regulator to continuously monitor compliance and risk positions being taken by actors in the market. Equally, a party could perform real time analytics of its risk position, the performance of its portfolio, or simulate future performance. This is all based on the real-time status of the contracts that a party holds, or is thinking of concluding. As each contract is explicitly defined it would become possible to trade on a defined set of contracts rather than an unknown bundle. For example, a party could trade explicitly on loan agreements made by companies in the oil and gas sector, based in Scandinavia, and providing the ability to take a greater risk position based on knowledge and insight it has about that sector. Finally, parties could use AI to perform continuous due diligence and allow them to evolve from algorithmic trading decisions to AI driven trading decisions.
My futuristic description in the paragraph above is clearly speculation of what the future may hold, but it is based on what is already technically possible. Some examples of each component of this diagram are set out at the bottom of this post.
Perhaps then the question should not be focused on what could the future be, but should perhaps be framed more as - "given that all of this is possible now, how do we make this or something similar happen?". The short answer to this question is that we need the whole ecosystem (including lawyers, clients, regulators and technologists) to develop a common vision, which in reality can only be achieved once all the actors in the ecosystem understand what is possible. I would argue that the value of speculating as I have done in this post is not in how accurately or not we can predict the future, but is more about testing perceptions to stimulate thinking about how different technologies can be combined and what could be created.
As a legal engineer, part of my role is to help educate people about the building blocks that make up 'possible futures' so that they can formulate their own vision of how their world could change for the better. To this end, we are developing a 'legal engineering capability model' at Wavelength, which we hope will help demystify the capabilities of different tools and solutions, and to move the conversation beyond the all too familiar "...I think we need some AI..."
Stay tuned - we will be exploring this subject further this over the next few posts.
Examples of what is already possible: here is a list setting out some examples of what is currently possible in the legal space using available technology, legal engineering and (design) thinking:
Structured documents contain elements that can be read by both machines and people. The data and entities, i.e. company names, dates, party roles, addresses, interest rates, currency value, etc., included in the document are explicitly defined using a controlled domain language such as FpML.
ClauseMatch is an example of a tool that could be used to create structured documents.
"A Smart Contract is a [piece of code] that executes terms of a contract" (Nick Szabo 1994). Simple examples would be a contactless payment or AXA's Fizzy, automatically executed flight delay insurance. However, the bigger goal is to automate more complex contractual activities as illustrated by the supply chain PoC which used IoT sensor data to dynamically price the value of cotton as it was transported from Texas to Qingdao.
Currently there are numerous competing 'smart contract platforms' of which Clause, Monax and Ethereum are examples.
Data enrichment is the ability to explicitly define the authoritative source for where a piece of information can be obtained and adding additional information, for example providing the Companies House number and linking to Companies House to add the company's name, directors, address, etc. From a computational perspective this can be achieved by linking an entity, company name, to the actual UK Companies House record via an API. This is an extension of the explicit defining of entities performed in a structured document and is central enabling more complex smart contract scenarios where a clause might be triggered in response to a real-world event. The open data movement has created a growing number of sources that could be used to provide computational definitions of things, i.e. Dbpedia (all concepts and facts from Wikipedia), Geonames (geographic repository of places), Permid.org (Thompson Reuters lookup services for companies and organisational structures).
Having created machine readable documents there needs to be a place to store them where they can be accessed and queried programmatically. There are many solutions that could fit this requirement from encrypted public blockchain solutions like Filecoin to secure/private content management solutions. The key being that this repository of information can be accessed and analysed by third parties.
Just as it is possible to create a machine-readable version of a contract (structured documents) and code to execute the terms of a structured document (smart contract) the same can be done for regulations. SmarT (aka Ganesha) from the GRCTC is an example of a tool that enables a lawyer to mark up a regulation using a controlled regulatory language so that a machine-readable model of the regulation is created in the background. A recent FCA TechSprint demonstrated that it is possible to create a machine-readable model of part of the FCA handbook and use this model to drive real time regulatory reporting by banks.
Text analytics makes it possible to create structured data from unstructured documents and for vast numbers of documents to be processed in this way. As a result, it is now practical to analysis a whole portfolio of documents and to visualise performance or risk across the portfolio as well as individual documents, for example Cognitiv+ and ThoughtRiver.
Portfolio analysis becomes even easier when considering structured documents and smart contracts as the need to perform text analytics is removed or at least significantly diminished.
Algorithmic trading is a well-established approach but more recently we are seeing the emergence of more sophisticated trading algorithms that incorporate machine learning, AI trading. This is a trend that will only grow as we become more sophisticated in the application of machine learning techniques and as smart contracts become increasing common.