News
Reveal Acquires Onna
Back to Content Library

eDiscovery Leaders Live: Ilan Sherr of DLA Piper

George Socha
March 26, 2021

Check how Reveal can help your business.

Schedule demo

Intro

Each week on eDiscovery Leaders Live, I chat with a leader in eDiscovery or related areas. Our guest on March 19 was Ilan Sherr, Legal Director at DLA Piper and Executive Director of Aiscension.

Ilan joined us to talk about the newly-launched Aiscension, a ground-breaking AI-enabled service designed to find cartel risks within corporations. Ilan, who is Aiscension’s Executive Director, worked with teams at DLA Piper and Reveal for four years to create this offering. Ilan covered their goals, the journey they took, and the results they have achieved. Ilan discussed the problems they sought to solve, their use of Reveal’s AI to tackle those problems, and the benefits corporations can enjoy by availing themselves of this first-to-market service.

Recorded live on March 19, 2021 | Transcription below

Note: This content has been edited and condensed for clarity.

George Socha:

Welcome to eDiscovery Leaders Live, hosted by ACEDS, and sponsored by Reveal. I am George Socha, Senior Vice President of Brand Awareness at Reveal. Each Friday morning at 11 am Eastern, I host an episode of eDiscovery Leaders Live where I get a chance to chat with luminaries in eDiscovery and related areas.

Past episodes are available on the Reveal website. Go to revealdata.com, select “Resources”, and then select “eDiscovery Leaders Live”.

My guest this week is Ilan Sherr. Ilan is Legal Director at DLA Piper and Executive Director at Aiscension. He is by training an EU and competition lawyer with an LLB from UCL, and then LLM and postgraduate diploma from King’s College London.

Ilan, welcome.

Ilan Sherr:

Hi, thank you for having me. How are you doing?

George Socha:

Doing well. Glad to have you here. This is a busy week for you.

The Launch of Aiscension for Cartel Detection

"The idea is that we can now help businesses spot cartels very quickly and easily, using the benefits of AI combined with the training and knowledge of DLA lawyers."

Ilan Sherr:

Yes, it's been crazy. You invite luminaries, but I’m very much the new kid on the block because our product only launched on Monday so it has been a busy week.

George Socha:

Well, it may have just launched on Monday, but it's quite the product. So, tell me, what's going on? What's the new offering, the new product?

Ilan Sherr:

At DLA Piper, we’ve just launched Ascension. That’s “AI scension”, so you can see what we’ve done there. It’s a cartel detection tool. What we've done is, we've been working with Reveal to create an artificial intelligence based algorithm that spots cartels within documents. The idea is that we can now help businesses spot cartels very quickly and easily, using the benefits of AI combined with the training and knowledge of DLA lawyers.

Cartels, Competition, and Antitrust

George Socha:

Let's roll that back a little bit. The problem of course is cartel, something…. Describe to us a little bit more, what that really means, what problem are you taking on here?

Ilan Sherr:

I should probably step right back for anybody who doesn't know what cartels are. Cartels are anticompetitive agreements between competitors to rig the market in some way, usually raising prices and damaging the market and it normally ends up with consumers paying more while having less competition or choices to choose. Cartels are a big issue for businesses and mainly because they’re illegal. They’re illegal in most of the jurisdiction in which we have offices, and most of the western world and lots of other countries are also bringing in new competition laws.

As you probably notice by the accent, I’m from the UK, and so we call it competition law within the UK but in the US you recognize it as antitrust laws.

George Socha:

It is essentially the same thing, competition and antitrust, just different labels?

Ilan Sherr:

Exactly, yes. Po-tay-to po-tah-to. The risks of cartels are quite high, and you can be fined massively by regulators in the US and Europe. The fines are up to 10 percent of global worldwide turnover of the entire group. In lots of countries it's also criminal, so there are criminal sanctions, your employees or directors could go to jail. if you have been found to be involved in cartels, then you’re exposed to damages actions, and in the US it’s triple damages. There are quite heavy impacts on businesses, and businesses want to be able to deal with that and prevent any problems going on in their business. The aim of this product is really trying to help those businesses in those situations.

Problems to Solve

George Socha:

I gather then, part of what that means is that they need to find out that they have a problem, initially, see if they can detect it, right? That's the first step in all of this? What is the traditional way of trying to accomplish that?

Ilan Sherr:

Normally you could internally have a whistleblower. Somebody could come up and say, “We've discovered this and there's been a problem”. That can be weeded out because you have new employees or an employee goes on training and realizes some of the things they’ve seen are not exactly sanctioned by the company or the business. Also, you can have a complainant, somebody who is alleging that you’ve done something wrong and as a business you’d like to respond and deal with that. Or the very worst end, you get the regulators knocking on your door, dawn raiding you and you find out that you have a problem whilst you are in the middle of an investigation. Normally you would respond that with an investigation.

I guess the question to businesses is, why are they finding out so late? And that's probably because the normal things that you would want to do to audit or protect yourselves have always been unreasonably expensive. The ability to do that if there wasn't a specific issue that you needed to deal with would not be justified by the price.

George Socha:

It would seem to me that if you've got a whistleblower, part of the challenge then is that you've got potentially a large volume of data that you need to go through. That sounds like an army of lawyers dispatched to go after a mountain of documents to try to find the few things that matter. That's assuming you've got a whistleblower or a regulatory agency comes knocking on your door.

I guess the other scenario is neither of those happens, but you're trying to put something in place proactive so that you can ferret out these issues before the whistleblower or the regulatory agency becomes a factor. Is that the way it works?

Ilan Sherr:

Yes, though actually that latter one is quite rare. The types of businesses that will proactively look for these sorts of problems of few and far between. Usually they are businesses that are either at a very high risk of cartels or, and this is more common, they've previously been found to have been involved in a cartel and as a result having been burnt once, having dealt with the issue, the management time that it costs to deal with it, the bad PR, and of course the fines and damages actions, they are willing to spend that sort of money and resources in order to prevent it from ever happening again. But, in our view, more businesses need to be in that position and proactively spot the problem and try to deal with it beforehand.

George Socha:

So, three issues: One, you don't find out the problem soon enough. Two, a mountain of documents. Three, an army of lawyers. Right?

Ilan Sherr:

Yes.

George Socha:

But you've got a different approach.

Ilan Sherr:

Yes. Yes. A long while back we came across that there had been a series of issues that have been happening in Europe where venture capitalist and private businesses had started being fined for the behavior of their subsidiaries. They invest in a business and they could be the 100% owner of the business or whatever. In the past, I'm not going to name the company, one of these private equity businesses, they had their investment company investigated by the regulator. They realized that their investment was going to take a hit because it seemed like they had been involved in the cartel, were going to be fined, and so they got ready to reduce the impact of their investments. But a year down the line as this investigation goes on, the investment company gets written to by the European Commission and they say, “Could you tell us what your global, worldwide turnover is?” And they get very upset and say, “Well, hang on, it wasn’t us. We weren’t involved in this. We knew nothing about it. It was our subsidiary”. And then the European Commission points to case law from the 1960s, saying that parent companies are liable for the behavior of their 100%-owned subsidiaries or any subsidy they happen to have control over.

They were fined jointly and severally for being involved with a cartel as well as their subsidiary. The problem is actually a very real one: how do you as an investment vehicle or how do you as a large conglomerate make sure that all parts of your businesses are compliant? It's been very hard, so we try to look at ways of how we solve this for a large number of our clients who want to be compliant with the law but because of the unreasonable costs trying to run these investigations haven’t been able to do that.

Pre-training eDiscovery AI

"we pre-trained it to spot these issues and then out of the box you press a button and the AI suddenly finds all of the potentially problematic documents for a lawyer to review."

We found a solution, actually in eDiscovery artificial intelligence. You can train an artificial intelligence just what it sees in the law, and we've been using that in cases to speed up our eDiscovery process and these tools are fantastic. I'm sure you have lots of other blogs talking about the results and the capabilities of those. But we came in from a slightly different angle which was, what if instead of training it on an individual case, what if we trained it on all the cases? What if we pre-trained it to spot these issues and then out of the box you press a button and the AI suddenly finds all of the potentially problematic documents for a lawyer to review. And then rather than lawyers using technology assisted review or predictive coding for finding something in an existing data set, actually the AI does the bulk of the work and brings forward and collects all the information for you. That's what we've done. That's Aiscension.

George Socha:

In terms of the cases you're using to train it on, I assume, although I don't know, that you are looking at things that have happened around the world. Not just in the UK, not just in the US. Am I correct in that assumption?

Why Focus on Cartels

Universality of the Offense

Ilan Sherr:

Yes. We chose cartels because I'm a competition lawyer, but also because one of the interesting things about cartels is it doesn't really matter where you are in the world, the cartel laws are pretty similar. There are a few other types of concepts within the legal profession where it doesn't really matter where you are, the laws are going to be pretty similar, but things like property and tax will be very different. The fact that the exact legislation might change but the behaviors you're looking for - competitors getting together to rig the market in some way, to fix prices, to carve the market between them or to rig bids - those behaviors are pretty universal, and so we can train that across all of our jurisdictions, get as much data as we can, and that all goes into it. The universality of the type of offense was quite helpful.

Lots of Decisions

The other interesting thing is that it's an area of law which is quite open, so there's lots of decisions on it. We've had access to the full decision jurisprudence of all the regulators across the world who published their decisions and then you've got the full-on cases in the court, so we had access to a large amount of public data. And then equally DLA is a global law firm, and we have competition practices in over 40 jurisdictions, so we also had access to a large number of internal data and examples that we could use anonymized in order to train it.

We’ve actually trained it on every single case since the creation of the European Union in Europe and we've trained in every single cartel case in the UK since the creation of the Competition Act and also in a number of other jurisdictions we've been adding that in. So a vast wealth of data, a huge amount of time that's the blood, sweat, and tears that goes into training an AI.

Many Years and Many People in the Making

George Socha:

It sounds like, then, an enormous breadth and depth of training involved. It also sounds like you didn't do all of this work in say, the last three weeks.

Ilan Sherr:

No. This has been in development for about four years and you kindly said, “You didn’t do the work”, and I can safely say, “Oh I didn't do all the work either”. I think over a hundred people at DLA Piper has been involved in collecting information, training it, and helping in some way in relation to the production of this. It's been a huge effort in incoming… Just finding all these data is quite hard, categorizing them.

I guess I should point out that we started off saying, “Can it support cartels?”, and we realized very quickly that actually that’s too broad a category for the state of the art that artificial intelligence is at. So we’ve traded spots, five different types of offenses: price fixing, bid rigging, market sharing, collective boycotts, and exchange of competitively sensitive information. There are other types of cartels, but it’s so rare we didn’t have enough data on them to train an algorithm so we’ll have to wait for there to be some cases.

George Socha:

Does that mean then that as you were doing this you constructed five separate models and trained on each of those five models as you moved forward? Or is it more complex than that?

Ilan Sherr:

As always it was more complex than that. We have a variety of models and the data scientists that we work with at Reveal have been absolutely fantastic in helping finesse and fine tune that. What we then did is we also tested against real data. That's when you really see the results and efficiencies that it creates.

Same Results, One Fifth the Price

George Socha:

What are those results? What are those efficiencies? What does it look like from that perspective?

Ilan Sherr:

The really interesting thing is that for the direct comparisons, in the largest case that we did, when we compared it against the, I call it “the traditional review”, the traditional review was using search terms, predictive coding, technology assisted review, and machine learning. So not people actually review documents. The AI and the alternative review found the same issues, but the critical difference was the AI and Aiscension - the AI and human review combined - managed to shave off about 80 percent of the costs. It was a fifth of the price.

Detect: Spotting Issues and Enabling Prevention

George Socha:

That's a considerable difference. What has the reception been so far? I know the announcement was only at the beginning of this week.

Ilan Sherr:

So, it's good fun, very interesting. Everybody we've spoken to about this is very excited. What's interesting is because of that cost saving, what you can now do. Obviously, we can run investigations in a totally different way, but we can also run and offer new products and services to clients that literally haven't been available before. They can now audit and detect problems.

I was talking about the issue that businesses have before in looking after their subsidiaries. A lot of the businesses that we see being involved in cartel conduct, all have very effective competition compliance programs. They train their employees, staff, directors, everybody involved in the business, what they can and can’t do.

But what's interesting, is still there are these businesses who do have compliance regimes, taking compliance very seriously, who are still being caught out. For me, the interesting part of that is the missing piece of the puzzle in a lot of these compliance programs, is actually detection, the ability to spot an issue. The way I always describe it is, when you go on your motorway or your autobahn, or whatever you call it in America, you see the speed sign telling you what the speed limit is so you know what you're able to travel at. And that's good, that’s like our compliance programs, teaching our staff the law, what you can and can’t do.

But the issue is what happens when people decide to get faster, and it’s very hard to actually police that. So how do you police that? Speed cameras. You put up speed cameras, or speed signs, so that people know that they're going above the speed limit. And when it's monitored, you get two behaviors coming out. One, you can find the problems and deal with them, and also, people when they know that there are speed cameras, slow down. And so the idea of adding detection into a compliance program means that you get the double-pronged effect of if there is an issue - someone unwittingly did something wrong - you're able to spot it and deal with it and deal with it before it becomes a problem. And also for your employees, they’re aware that they are going to be audited and as a result will be a bit more careful and wary and the issues will come to their mind. Prevention is the best cure.

Enabling Audit of Today’s Massive Volumes of Data

George Socha:

And I suppose on that, “they are aware that they're going to be audited” side of things, a big difference is that in the past, yes they might be audited but the chances of that actually happening were pretty low because it would be such an expensive and time consuming process. And now the chances of being audited are much greater, there's a camera every block instead of a camera every city. Is that right?

Ilan Sherr:

Yeah. It brings it into the realms of possibility. I think there will be a number of businesses that would audit this type of activity, but I think the majority of businesses that we come across, they will have compliance programs and regimes in place. It's very hard to audit. The exponential growth of data has made it much harder. Twenty years ago if you were to run an audit, you could look through someone’s files, you could run through their diary, their commentary, their notebooks, it would be quite quick and easy to spot. Businesses over the last 20 years have become much larger conglomerates. Businesses are growing. There are more people across more offices. It’s very hard to monitor people across many jurisdictions with different rules applying to them. They're disparate, it is hard to physically go there. And equally the explosion of data, the amount of data that we're creating on a daily basis has just exploded. So, the volume of data you need to go through, the communications, the emails, and the chat, all of that, it's very hard to stay on top of. Each of those are risk areas of the business. Our eDiscovery tools have enabled us to do that but the costs of that have not scaled with the massive scale of that exponential increase in data. I think AI is the real solution that comes in and is able to do that, it automates that process. And it’s very effective in the way it automates that process. It’s brilliant.

First to Market

George Socha:

Well clearly this was not, as you said before, this was not something you rolled out overnight but it’s years in the making. I don't even know how many hours of effort must have gone into this. Are you first to market with this? Is there anybody else out there that has something comparable to this?

Ilan Sherr:

We’ve been monitoring the market. As far as we are aware, there is nobody else that has done anything like this. We are first to market. To be honest it’s a bit of a surprise because for us it feels like a no brainer, it's so obvious that this is what you can do with this technology. It is so obvious that this is what the need is and certainly a lot of the businesses that we've spoken to have been very excited about the opportunity and what this could mean for them and how they could work out. It's nice to be first to market and also it's a great opportunity for us and we can help clients as a result. So, win-win.

The First Product: Using Aiscension to Audit Cartels

George Socha:

Win-win sounds great. I’ll ask, you decide whether you want to answer. Where from here? What next?

Ilan Sherr:

Now it’s the easy part. Sales.

George Socha:

Okay.

Ilan Sherr:

Yes, so it's good to go. The first product that we're launching is effectively an audit product, it will audit the cartels. The way that works is, a business would identify their risk area, what it is that they want to investigate and check or audit. They would collect that information, so they identify, usually the individuals within the business who are within that high-risk category that may need to be audited. They then look at where they hold data, where is it stored, does the employer have the right to access that data, and if they do, then that can be harvested. Rather than taking a forensic capture because this is an audit, the businesses can do that themselves and that saves them a massive amount of cost. They collect that. We’ll upload it onto the cloud, we then process it, very soon it’s eDiscovery at this point. We train the Aiscension and AI tools at it, which will then effectively rank the documents in relation to the levels of signals of cartel activity.

The data scientists at Reveal have been amazing. Dr. Matveeva and Isha Chopra, they’ve created an expiration queue. That takes the best examples of each cluster of documents so that we aren't repeating some of the work.

Then the lawyers will review the cream, the top sample of documents, in order to see whether there is an issue and provide a report back. That’s how it works, that's what we’re launching now, and that's what's available.

What’s Up Next

But as you can see, there are a lot more things that we can do with this. The monitoring functions that Reveal have already created, we're looking at rolling that out also in the future. That will allow us to actually monitor every new communication sent, every new document created. That can really allow businesses in real time to proactively spot and stop an issue before it becomes a problem.

The Benefits of the AI

"I think very quickly the AI will have been able to review more documents than a human would be able to review in their lifetime."

George Socha:

Quite amazing. Quite the undertaking. Congratulations. Is there anything that I have neglected to ask you about this initiative that I should be probing about here?

Ilan Sherr:

I guess the benefits of the AI are quite interesting. What we found is in order to run some of these reviews, we needed to staff it with large teams of lawyers, paralegals, trainees in order to churn through the volume of data, even after using technology assisted review or filling out the data and removing noise or duplicates and the like.

What we find is there's a lot of inconsistency that happens. At the extreme polar ends of “smoking gun cartel” to “totally non-proprietary document”, people tend to get to the same results. But the areas in the middle, the gray areas, that's where you find a lot of inconsistency.

The benefits of having an AI are twofold. One, you have a very consistent result. The AI comes to the same conclusion in relation to all the documents in the data set. You have a very clear, provable methodology of what's been done and you can follow that through, which is great.

But the other thing is, obviously, it takes a lot of people to train it and it’s trained over time. You don't lose that group capability, you get group think in the creation of the AI to develop it, train it, and refine it. But you still have that one result at the end. So it's very interesting.

Our senior partner Simon Levine also pointed out that the other benefit of the AI is you don't need to give it a performance management review. At this time of year, we are all going through our PMRs. You save a bit. I mean, we do do some quality control, but you’ve got to do that with lawyers anyway. Also, it doesn't sleep, it doesn't take breaks or drink coffee, and also its mind doesn't lapse, so it's very interesting.

I think very quickly the AI will have been able to review more documents than a human would be able to review in their lifetime.

That's the massive benefits you get of throwing something like this at a problem. Lawyers are safe, it can’t be an analysis just yet, so we still have a job with the fun work of the analysis at the end so we can sleep well at night so far.

George Socha:

Right. So, once again, it's not John Henry versus the steam drill. It's not technology replacing all of us, it's the super exoskeleton.

Ilan Sherr:

Yes, for us what's clear is actually what it's allowed us to do is (a) be more competitive in what we can offer in relation to responding to cartel investigations.

But the critical thing is, now we can offer new services that were just not feasible before this was available and the cost would have been prohibitive. That allows clients and businesses to be proactive, to detect the issue and prevent it. The costs of prevention are just so much better than the costs of cure. It just makes perfect natural sense. And I think as a lawyer, it's nice to be on that side of the fence. Professor Susskind puts it as “the fence at the top of the cliff, rather than the ambulance at the bottom”. I think it’s a more comfortable position to be with the businesses that we’d like to help.

George Socha:

Very much so. Ilan, thank you very much for joining us this week. My guest this week has been Ilan Sherr. Ilan is both Legal Director at DLA Piper and the Executive Director of the brand-new, just-announced Aiscension. Ilan, thank you very much for joining us.

Ilan Sherr:

It’s been a great pleasure.

George Socha:

Our guest next week, March 26th will be David Yerich from UnitedHealth Group. Ilan, thank you once again.

Ilan Sherr:

Thank you.

Get exclusive AI & eDiscovery
insights in your inbox

I confirm that I have read Reveal’s Privacy Policy and agree with it.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.