During The AI Summit in London, AI Business got the chance to speak to Jan Van Hoecke, Co-founder and CTO of RAVN Systems. RAVN are one of the leading AI solution providers for the legal sector, and were shortlisted for Best Enterprise Application of AI at the AIconics Awards.
We caught up with Jan to find out RAVN’s story, how they broke into the legal market, and a bit about their plans for the future.
Jan Van Hoecke of RAVN Systems
Back in 2010, Jan and his fellow RAVN co-founders struck upon a key question: How can we get machines to interpret unstructured data?
Jan explains: “In the financial sector there is massive amounts of structured data, in Excel spreadsheets, but machines here have an easy mission. It’s easy to put in a rule, or pick out a cell by matching a row and a column – this kind of data is very compatible with machines. On the other hand, text is unstructured data that is very compatible with humans, but totally incompatible with machines – machines can transmit text but they cannot interpret meaning from it. If in a contract in full text it says, for instance, ‘the value of this contract is £25million on the condition that x, y and z happens’, it’s very difficult for a computer to understand what it means”.
Jan and his team decided this was something they needed to tackle. Going beyond corporate keyword search engines, the RAVN Applied Cognitive Engine (RAVN ACE) is able to read, interpret and extract specific information from documents, just like a human would do: “the machine not only has to pick the right document, it also has to go into each document and pick out the right information from it – that’s step one”, Jan says. “Step two is then to dissect the document – pull it apart and get meaning out of it so you get an answer to your questions. That’s the high-level concept.”
Picking out information from documents (i.e. step one) proved a valuable internal tool, and in this RAVN’s solution found some success in the enterprise. But its revolutionary functionality – the ability to extract meaning and get answers to questions – took a little while to find its way into the market.
Last spring, RAVN got their breakthrough in the legal sector with international firm Berwin Leighton Paisner (BLP).
Jan explains: “We thought about the industries where thousands of documents are read through by humans, and we came to law. If you can eliminate lawyers, which are expensive, from reading and interpreting legal documents such as contracts, you save a lot of money, so you can quantify the business case there. These lawyers can then focus on more interesting billable tasks.
“When reading a contract, the system will recognise whether it’s the title page we’re looking at, and say ‘this is the parties section’, etc. It will then start to process the blocks of text using Natural Language Processing. Then within that block of text it recognises the company name, the company reference number etc, which enables it to get to the meaning of the sentence. It then reattaches the meaning to the document as metadata in our system.”
“We did a trial with BLP, and the results produced were amazing, but naturally they did a QA check on everything before signing up fully. When the calculator first came out, my dad told me how he used to check every answer that came out of it – now we would never imagine doing that.”
Once the system was fully implemented at BLP, the story got picked up by BBC News. “The demand was higher after that!” Jan recalls.
So where next for RAVN?
Jan and the team have an avenue in mind: “Our mission for the future is to break down the barriers and make RAVN’s Artificial Intelligence available for all verticals across the globe and surface its true multipurpose capability. We are already working in the financial sector on multiple cases, including extracting key terms from ISDA CSAs. We see RAVN ACE being used from large organisations, bring improvements to everyday life by applying it in IoT applications, and in general powering systems and tools that can respond to new and uncertain questions quickly with intelligence.”
“The next thing is prediction. It’s one thing to extract information, but it’s another thing to predict outcomes. This could be very valuable in the legal profession, especially in small insurance claims for example, where you can assess whether it’s worth making a claim or not.”
In the future, Jan suggests this could extend to court rulings at the top level:
“AI can consider a lot more information than humans – if the pattern is in the data, AI can probably predict it. The question is: do you have the right data, and are the parameters correct?”
We spoke to Jan at the inaugural AI Summit in London on 5 May. The second, larger AI Summit takes place in San Francisco on 28-29 September. To find out more, and to join us at the Fort Mason Center in September, visit: theaisummit.com
Image credit: https://www.ravn.co.uk/company/meet-team/