A Singapore-based company is aiming to cut the time it takes big banks to conduct investigations on suspicious trade finance transactions.
Most banks will have transaction and customer screening solutions in place, but Silent Eight founder Martin Markiewicz says that his company’s technology goes much deeper than solutions already on the market. The company uses machine learning and artificial intelligence (AI) to monitor transactions, flag those that are suspicious, and outline potential scenarios.
Rather than sifting through the data themselves, analysts are presented with a dossier of information about suspicious transactions within a minute.
The system, he tells GTR is trained to work and think as an analyst in a bank and does the work in a fraction of the time. Already, it has been picked up by a number of banks. Singapore’s OCBC has gone public with its piloting of the solution, while Markiewicz says that it is in the process of completing a proof of concept (PoC) with “one of the biggest banks of the world” that will show the software’s capabilities in trade finance.
We spoke to him to find out more about the company, the solution and where he sees it fitting into the trade finance industry.
GTR: What is Silent Eight and how does it work?
Markiewicz: We work with big banks and bring a fresh view to the whole compliance part of transaction monitoring and customer screening. We’ve built up a solution to help them do this more efficiently and accurately.
In a bank’s transaction monitoring process, they’re trying to figure out which transactions are suspicious and then want to investigate them. Usually what’s used by the industry is a rule-based system: you decide what you define as a suspicious transaction, you write a few scenarios to define these. You may be customer screening and wanting to avoid dealing with sanction-linked entities. Every time you’re doing these activities you want to look as deep as possible, you don’t want to miss something. But as you look deeper and deeper, it creates more and more work – sometimes more than you can handle.
Our system is trained to work like an analyst. Instead of passing it to a huge army of people, you pass it to our engine first, and our engine solves these cases. Only very suspicious ones are then passed for human intervention. We’re making this whole process scalable.
As a bank, you should look at every transaction as a potentially suspicious one, it should not depend on the size. Criminals are smart and will dodge the way your systems work. But if you inspect every possible outcome using this technology, you are saving on a whole team of analysts. We’re saying: if we use AI in this space, train our engine to work like your analysts, doing the work of an army of analysts over months in seconds, then leave it to analysts to review, it will be a scalable solution.
GTR: Where does it fit into the trade finance industry?
Markiewicz: The reason is the same for trade finance as for transaction banking. A lot of things in the bank have to be checked and investigated. In all the banks we work with, this process is heavily manual. If something gets flagged for investigation, an investigator sits down and does it manually. It depends on the case, but it can last for a very long time. You need a lot of people and the more investigations you do, the more you need.
We train our engine to work like an investigator, it will solve a case that would take an hour to crack, in a few seconds.
GTR: How does the risk modelling and data pulling process work?
Markiewicz: First of all, we sit down with the analysts and talk for a day or two to figure out where their data is coming from. We map the data sources and plug these patterns into our system. This is done by machine learning. We look at what the team of analysts was doing in the past and learn from that how the engine can behave according to their team standards or rules.
We don’t come up with those rules, we learn them from the bank. The analysts advise on the engine we put in the bank, then it learns how to operate within the bank. The end effect is something unique to every bank, because every bank will have a different set of data sources, of handling situations. We put the solution inside the bank, map the data sources and train it to work like one of the analysts from the bank’s team.
The process will obtain information from multiple systems and data sources – often not even inside the bank – also using search engines to fill in the gap, and then when the whole picture is ready, the conclusion is drawn using AI.
If a company is flagged for investigation, you would have to look at everything they’ve done for the past few months: all the counterparties they’re dealing with, all the directors and employees… to make sure nothing bad happened. Our engine will do what it does, this investigator’s job, automatically. It pulls all the evidence from inside and outside the bank in the public space and reasons like an analyst to come up with an explanation as to what happened.
GTR: In Singapore is there much competition in the regtech and AI space for trade?
Markiewicz: I don’t look at this from a Singapore point of view, but globally. Right now we’re scaling to match the global reach of the banking customers we have. When I look at the landscape right now, it’s not a crowded place. A lot of people want to try something, but I don’t see that many players who can tell banks: ‘Wwe can help you reduce your cost of doing processes by 50-80%,’ and then deliver.
We’ve found a nice spot for ourselves and we can deliver on our promises. I don’t see us being in a market like selling databases or cloud computing, where there’s a few dominant players. The scenario in our space is that there’s obviously a lot of people doing screening engines, and these engines are flagging up things to investigate. We’re not competing with them, we say: keep the screening engines you have now, but lower your threshold and allow more suspicious things to be flagged up, but instead of hiring more people to investigate this manually, automate the investigation part.
GTR: What is the next step in getting the solution market ready for trade finance?
Markiewicz: We’re finishing a PoC with one of the biggest global banks. The idea of the PoC is just to prove the business case, and to make sure the amount of money we’re charging for a license makes sense. We’re not testing if it works: the PoC will prove that we can reduce manual work by huge factors. Looking to the future, banks are going to have to be more thorough in investigations, hiring more people, costing more money. You won’t have to do all these when the regulator raises the bar if you have a system in place.