Baker McKenzie has entered into partnership with AI research engine SparkBeyond in a bid, ultimately, to automate legal processes, vastly expand productised client offerings, and if the ‘holy grail’ is achieved, create what is described as “a problem solving operating system that sits with knowledge workers.”

SparkBeyond, which has been championed by Microsoft CEO Satya Nadella, crawls client and public data to test millions of ideas and generate millions of hypotheses to identify patterns with the most actionable impact. It works with companies including MetLife, Anheuser-Busch, and McKinsey & Company, and says it has had a $1bn impact on the companies it has worked with.

The partnership is being run out of Bakers’ global innovation arm Reinvent, which is headed by partner and head of R&D, Ben Allgrove. Allgrove told us: “Like other law firms we have deployed, to the extent that it’s out there, machine learning to improve some fairly narrow applications based tasks and machine learning is getting better but it doesn’t change the job or business model or how we work for clients.”

He says that Bakers has been looking for several years for a way to apply machine learning more horizontally but couldn’t find the right provider, commenting: “What was clear with big brand names was that unless you turn the tap on and give them a multi-million contract up front, they are not interested. But with some of the startups, they don’t have the resources to deal with the big firms.”

About a year ago Bakers started speaking to SparkBeyond, which was founded in 2013 in Israel by Sagie Davidovich and Ron Karidi, and Allgrove says: “We were interested because they are taking a significantly different tech approach with their hypothesis engine and using humans to test their hypotheses at scale. It gives you options not answers, which we thought was more scalable.

“Their entire model is based on the glass box, so explainability is built into the technology architecture. In concrete terms, if you ask their engine, as we are, ‘what are the key drivers of client demand’, they will say ‘here are 10 suggestions’, and we hope that thrown up are ones we haven’t already thought of.” Glass box AI models are the opposite of black box models, in which it is far harder to understand and explain how an output is reached.

While AI projects are often held back ironing out issues with data, Allgrove says: “SparkBeyond don’t spend a lot of the time structuring the date. They will take data from all our internal systems, from financial to knowledge, all GDPR compliant, and they will combine our data with what is publicly available to generate answers to the questions you are trying to solve.

There are four stages to the partnership, although Allgrove candidly says: “Whether we get to the fourth I’ve no idea.”

Stage zero was finding the right partner. Stage one is to prove that the tech can work in the legal space, and here Allgrove says: “Our initial project is internal facing quite deliberately.” Allgrove needs to convince his stakeholders that this project is worth the investment.

On the basis that it is, stage two will bring in clients and involve a significant allocation of resource. Allgrove says: “The second stage will be client facing apps, where we say, ‘here is what we can see, do you want to plug in your data?” Baker McKenzie could, for example, look at clients’ risk from antitrust investigations because of signals in the market. SparkBeyond’s platform could also be used to identify non-legal risk in M&A to assist the due diligence process. Or it could be used to, in Allgrove’s words, “turbo charge trademark searches.” To do that properly Baker McKenzie will need to, no pun intended, bake in client data.

If stage two is successful, stage three will be to productise those client offerings. Allgrove says: “We will take the learning and see which can be made into a horizontal platform. If we discover that something worked for one, two, three clients, can you build a platform for 1000 and productise it?”

Stage four, the most ambitious, is to build a problem-solving operating system; in effect an environment for knowledge workers that sits with them. This, Allgrove describes as the ‘holy grail’ and says: “Stage four is about putting the technology where the humans are and you won’t be seeing an announcement about that for the next three years.”

What will be happening in the background during that time is perhaps one of the most interesting plans of this partnership.

The issue with RPA or ‘law at scale’ when it comes to high value legal advice is that everyone works in different ways. There has been very little success in automating anything but high-volume legal processes because it’s expensive and timing consuming and change management is incredibly hard. Allgrove says of the new partnership: “We will start with one practice where the machine is in the background and says ‘I’m watching Caroline, Jane and John. We want an environment where the machine says ‘both of these are the same but done in a different way and I can understand how they can be the same.’ It can take a piece of what you’re doing and make it better over time.”

Reaction to the news has been mixed, with one vendor commenting: “This reads a little like a press release of yonder year about AI – it’s going to do this AI magical thing and predict what customers want. Very holy grail-esque. Light on the details and big on the promises.”

However, Hyperscale Group and LITIG founder Derek Southall said: “This looks like a really interesting move. Admittedly this is only the start of the process but  partnering with a well-regarded player like this in this area rather than trying to do it themselves could deliver a step change in the professionalism of outputs – Hogan Lovells have been open about partnering in the past and many larger firms do this with ALSPs. Baker McKenzie will obviously have domain experience in spades to bring to this too.”

He added: “What I really like is two things – first the fact there is no specific agenda. They seem willing to initially experiment and explore as opposed to needing an immediate ROI which is a positive. Secondly Baker McKenzie are obviously one of the truly international firms. If their global outlook and footprint is properly leveraged we may see some very different products emerge. One to watch.”

Respected legal consultant Tony Williams, principal of Jomati Consultants and former managing partner of Clifford Chance added: “The Bakers move is interesting as it involves non-legal sector technology being used in the legal market. Much legal tech appears to be relatively introspective ie solving a specific legal issue rather than reimagining the issue from the clients perspective.”

Much is riding on the partnership with SparkBeyond living up to its expectations. The company has strong ambitions for its technology to be used as a force for good, which Allgrove says resonates with the firm and is a big reason why their values align. He says: “At my first meeting we talked about social impact and the founders at SparkBeyond are very keen that their technology has a good social impact. The opportunity to deploy it to increase access to justice appeals to them and resonated with me. One of the big reasons for going with them is that our values align.”

He adds: “My view is that one of the reasons we haven’t seen fundamental disruption in law is that the ROI doesn’t stack up. Vendors want $50m up front. The approach here is to find someone smaller and willing to invest with us in this project. We’re offering a channel into the legal market and an ability to create meaningful change.”

caroline.hill@legalitlexicon.com