Serious Fraud Office

Each case requires SFO investigators to review staggering numbers of documents. “The volume of data is a huge challenge,” said Day. Ben Denison, chief technology officer at the Serious Fraud Office, described some recent examples. “The Panama Papers leak involved 10.5 million documents. Our investigation into a bribery case with Rolls-Royce had 30 million. We’ve got one now with 60 million and another one in the pipeline with potentially 100 million documents.”

To meet these challenges, the SFO has turned to artificial intelligence (AI), a powerful new weapon in the fight against crime. Using AI to review, analyze and provide insight into vast amounts of digital content, the SFO can expedite its investigations and improve its chances of successfully prosecuting economic crimes.

“We’ve had to change the way that we work. We can no longer just sit in front of a data pool and throw search terms at it,” explained Day. “It would take a team of 10 reviewers three years to look through two million documents. It’s just not possible. We have to find defensible ways to cut it down. AI technology is no longer ‘nice to have’ for us—we have to have it.”

When the time came to replace its older generation eDiscovery technology, the SFO prioritized its requirements, with user experience at the top of the list. Denison described the selection process: “The case teams, document reviewers, investigative lawyers, forensic accountants and eDiscovery support people— all of these users were part of the bid review process. Much of the overall decision was based on user feedback, which is absolutely critical because you want to base the decision on whether the solution is really going to make an impact to the business.” Day added, “There was a lot of user testing, a lot of requirements and OpenText™ Axcelerate™ came out at the top.”

OpenText Axcelerate is an eDiscovery and investigations platform that uses advanced analytics and machine learning within a fully integrated, intuitive review interface. Axcelerate pioneered predictive coding for legal review and features a continuous learning model that is flexible enough to accommodate unpredictable, rolling data loads and evolving case theories. In addition to AI, Axcelerate integrates a variety of analytics and investigation tools that have helped Day and SFO investigators understand the story behind the data.

“We use OpenText Axcelerate to get the data to talk to us, with prepopulated metadata fields, analytics and visualization tools,” explained Day. “All those things very quickly help us understand where our data is. And, being able to do things quicker and better was at the heart of everything we wanted from a document review platform.” OpenText Axcelerate is now being used for all new cases at the SFO, and the early results are promising. “We’re presenting OpenText Axcelerate to investigators to use the tools and analytics to get to the documents and to understand the case quicker. That’s really important to us. OpenText Axcelerate changes the way that we investigate. It has become a much more intuitive and iterative process. We’re embedding that in the culture of our investigations now,” said Day.

Axcelerate automatically recognizes patterns, groups information by subject and organizes timelines, with features such as predictive smart filters, phrase analysis, data visualization and communication mapping. The solution helps prioritize the review of more than 100,000 documents a day, which is impressive compared to human review limitations of just a few hundred documents per day. According to Denison, “The move to OpenText Axcelerate is a real sea-change in terms of the technology. Compared to the old system, the OpenText infrastructure is far more scalable and can take on far more data and analyze it in a much more efficient way. The solution is also more efficient in terms of user experience. It’s going to have a huge impact on our investigations.”

Denison commented on the beneficial features of the solution. “OpenText Axcelerate’s advanced analytical tools are very useful when you’re dealing with cases that could easily have tens of millions of documents.

With phrase analysis, for example, we can identify a term that an organization might be using to refer to bribes.” Another feature is the concept browser, which uses unsupervised machine learning to organize documents with similar contextual themes, even if they don’t necessarily use the same keywords. “For example, we might think we’re looking at behavior in a particular country, between a particular group of people in a particular timeframe. Using OpenText Axcelerate, we can narrow down what we’re looking at and see what they’re communicating about,” explained Denison.

With this new weapon in its arsenal, Denison predicts SFO staff will be that much closer to cracking the case. “By automating and streamlining the simple parts of the work, we allow investigators to use their brains more, to do the harder work, the more complex work, the work they enjoy.” Day added, “The people who use OpenText Axcelerate every day love using it.”

Day acknowledged the value of OpenText’s support in deploying the new solution. “We don’t see OpenText as just a supplier—this is a close partnership.” Denison also commented, “Having people who really understand the product, and who’ve been working with it for years, is a huge benefit.”

In its mission to bring criminals to justice, the SFO looks forward to accelerating investigations and prosecutions with this new technology. Day explained, “Bribery and fraud are particularly pernicious crimes that ruin lives. To help rid the world of that is important to me. It’s important to all my colleagues at the SFO.” Denison concluded, “OpenText Axcelerate’s advanced analytical capabilities are absolutely crucial to how we investigate our cases in the future and to make sure that we do so quickly and efficiently."