Bad news for crooks: RAVN’s AI speeds up the document-sifting process – and is more accurate than humans
The Serious Fraud Office (SFO) had a problem. Its investigation into corruption at Rolls-Royce was inching towards a conclusion, but four years of digging had produced 30 million documents. These needed to be sorted into “privileged” and “non-privileged”, a legal requirement that involves paying junior barristers to do months of repetitive paperwork. “We needed a way that was faster,” says Ben Denison, chief technology officer at the SFO. So, in January 2016, he started working with RAVN.
Pronounced “Raven”, the London startupbuilds robots that sift and sort data, not only neatly presented material, but also unstructured documents. “Where someone has scanned 300 pages, it’s not uncommon to put one page in upside down,” says co-founder Peter Wallqvist. “We need to deal with that real world of messy datasets.”
The two teams started to feed material from the Rolls-Royce case into the AI. By July they had a viable system, and with the agreement of lawyers on both sides, they set the robot to work. The barristers were wading through 3,000 documents a day. RAVN processed 600,000 daily, at a cost of £50,000 – with fewer errors than the lawyers. “It cut out 80 per cent of the work,” says Denison. “It also saved us a lot of money.” For Rolls-Royce, it had the opposite effect. In January 2017, the engineering company admitted to “vast, endemic” bribery and paid a £671 million fine. “It’s hard to imagine a better outcome,” says Wallqvist.
RAVN’s co-founders – Jan Van Hoecke, Simon Pecovnik, Sjoerd Smeets and Wallqvist – met at Autonomy, the UK’s first unicorn, where they worked on early versions of AI-powered database management. In 2010, the four left to launch RAVN. The self-funded firm now has 51 employees, revenues of £3 million and around 70 clients, mainly city law firms. BT, which signed a “very significant” deal, credits RAVN with annual savings of £100 million, due to automated checks that ensure contracts’ accuracy.
Plus, of course, there’s the SFO, which is using RAVN in increasingly clever ways. That means allowing it to make subjective judgements, including pointing investigators to data it thinks is relevant to a case. “This is potentially very valuable,” says Denison.
Wallqvist believes the system can go even further and make not just assessments, but predictions. For example, by suggesting likely outcomes of mergers and acquisitions. “We’ve gone to the level of figuring out and structuring data,” says Wallqvist. “Now we have the ability to surface that record of the past to predict the future.” Today, Watson. Tomorrow, Holmes.