Before I begin, I did just want to cover what Tara actually entails. I’ve mentioned her skill before, and so it really is around payment sanction screening alert review. So dispositioning those alerts, and it’s done in a few ways.
Included in that is an automated business process that contains models that allow for entity recognition. So being able to understand what is an individual versus a company versus a vessel versus some other entity. Being able to do name matching and not just exact matching, but using true machine learning algorithms to determine likeness of names and determining if it really is a match similar to how a person would. As well as the ability to use a decision matrix and historical decisions to improve on the decision making, moving forward. Sitting on top of all of that, we have analytics, you can actually do a performance review of Tara to make sure she’s meeting KPIs. As well as we include integrations and out of the box points of contact with your existing systems. So what we’re going to see in this demonstration is how we would interact with Fircosoft, but we also have the ability to integrate with other payment sanction screening tools. We can talk through that, but I’ll show you just one example.
Just to kick things off, the way that teams are working now is they’re working within a system. What they’re seeing then, is a list of results that are returned based on the details of their transactions, what things have hit against watch list entities. They get a list back, they have to go into each one, research each result, see if it’s a real match or not, or if it’s a false positive, mark it, document it, close it and move on.
Where Tara fits in with the existing team, is that she’s actually able to take that work upfront, analyze all of the results, and what gets returned back to your analyst team then, is oftentimes a smaller list because the false positives are removed, but you can also see that there are things that are half passed. Tara has done some of the work up to a certain point, and now it’s for time for an additional review on top of the work that Tara has done. It allows your analyst to come in and focus on work that is higher value, and eliminates a lot of that back and forth and manual looking up information. It’s all going to be presented to a person in one view.
If we look at the details inside of Fircosoft, for example, what you can see here is again, the payment details on the left hand side. So this is the information within your system, and your payment screening, the results from the watch list entity. Then you can actually see that Tara has entered in information into Fircosoft directly that says, “Hey, this is a false positive. This is not actually a geographic match.” Just like a person come can come in and say, “Oh, this is actually in Turkey, but the watch list entity is in Syria. That’s not a real match. We’ve marked it as a false positive. And we can continue on with the next alert in the system.” This then unlocks the payment, so that your customers are able to conduct their transactions.
What I’m going to talk through now is how Tara makes that decision. It’s a black screen now, but what I’m going to show you is some examples, walking through exactly how Tara is able to determine if something is a false positive or not, or if something needs additional review. This view that I’m going to show you is a manual task, that likely your teams would never interact with. Your team is still working in the system, but this is just giving you an example of how WorkFusion is working in the background. The first example here is again, showing on the left hand pane, this is the payment information. Here is the information that was returned from the watch list entity that caused this to be an alert. And then we can actually see that Tara is doing the work to find information within that payment message.
In this case, we can see that this actually hit on Iran. But if we look at the actual text, Tara was able to pull out an address, a country, and a person, and actually understand that this is not an actual match. The watch list entity hit against a fragment of pieces of data. There’s two different entities, but there’s the text “Iran” that matches. But actually it’s part of an address, and an individual name so that actually does not match the watch list country. That’s how Tara made the decision, in this example. Tara is using machine learning models to extract the pieces of information from that SWIFT message, or your own description in your proprietary system, and comparing it truly against the watch list entity without using keywords or exact matches. So just like a person makes decisions.
The next example is going to show here that again, we’ve got the payment information on the left hand side, we’ve got the watch list entity hit information in the middle. And again, we have Tara pulling information from that payment on the right hand side. As you can see that we’ve identified an address, so this is a piece of information that we’re pulling from that payment information. In addition to comparing what’s in the payment, and what’s in the watch list and entity hit is we can actually see, “Hey, this location is not a match. We actually verified. We went out to a third-party source to validate that this is a true address. And it’s in Virginia in the United States, and actually is not the embargoed country.” So not only making a decision on what information is in the payment and the alert, but going out and doing additional information to confirm that they are making the right decision.
The next example is very similar. Again, Tara is able to view the payment information, extract specific pieces of information out here. So again, we’re able to extract information from that free text. We’re able to identify what type of information is in that field. In this case, we can see an address, a country, a company, and so on. So just like a person, I can go out and look up in third-party websites to validate information that might be needed for me to make this decision, Tara is able to do the exact same thing. In this example, we can see that this is actually a false positive. The screen’s entity is located in France. And the watch list entity is actually in Syria. So we went out, we validated information, compared it to the results, and Tara was able to make the decision that this was a false positive. Again, Tara is able to take the information from the payment details, from the watch list entity details, and do comparisons.
This comparison is using machine learning, it is not just using an exact match or a fuzzy match. Some examples of what this would be is we’ve seen address and location being something that we’ve determined as false positives. A location may match if they’re actually neighboring states or countries, so that’s something that we take into consideration even when it’s not an exact match. We’re able to compare entity types, so something as an individual, versus a company, versus a vessel. Tara can actually determine that and determine if something is a false positive based on that. Then I mentioned the name matcher, right? So it’s not necessarily an exact match, we could take things into consideration like nicknames and similarities of the match as well.
Also included is, once Tara has done her work, how do you tell that she’s doing job that she was hired to do? Just like everybody has KPIs in their day to day work, we can use the analytics provided out of the box with Tara to see and conduct her performance review. We can see things like, What is the alert rate for false positives? Of the alerts that have been generated, how many did Tara determine were actually false positives versus those that she took no decision on and maybe escalated to the rest of your team? And then for those that were false positive, why actually were they determined to be false positives or not? So you can see actually the reasoning behind this, because Tara not only makes a decision, she actually documents it and passes that back into Fircosoft or your tool of choice, so that you can document the reasons why these things were considered false positives. So her decision is included in that.