Avoiding fraud in real time through artificial intelligence tools is the new commitment of IBM with banks and other types of institutions, especially now that digital transactions have exploded after the Covid-19 pandemic.
According to a report on the global impact of financial fraud that this technology company developed in conjunction with Morning Consultant, fraud detection models are only executed in less than 10% of high-volume transactions. In addition to this, according to data from the National Commission for the Protection and Defense of Users of Financial Services (Condusef)customers can take up to more than 100 days to realize the contracting of unrecognized products and identity theft in their bank accounts.
Baltazar Rodríguez, senior architect at IBM Technology explains that normally the fraud was attacked after the transaction had started.
“That is: these transactions were received and passed through an analysis mechanism to determine where there was fraud or not. The problem is that these analysis mechanisms could not normally be done in real time, they were done after the transaction. For example, a message was launched to say: I just registered a transaction on your card that seems atypical to me, is it true or not? (…) the bank is going to do its best to try to finish the transaction, ”he specifies in an interview.
However, it complements, in some cases, and especially in digital transactions, the analysis is done after the execution, and then it is difficult to stop the fraud. “Then the mechanism that you have to use is a compensatory one, such as having an anti-fraud insurance,” he points out.
Processor with artificial intelligence models
In this sense, Baltazar Rodríguez mentions that IBM developed a new processor, which has all the capabilities to support large volumes of transactions, as it already existed, but with an additional one specialized in the execution of artificial intelligence models.
“We connect it in such a way that those transactions can go through artificial intelligence models in real time (…) now what I can do is take, check and see that you actually have funds, but additionally pass all the transaction data to through an artificial intelligence model, and thereby determine, for example, if the place where you carry out the transaction is where it is normally done or not; if it is an atypical transaction, at an atypical time or location (…) and detect in real time if it has a high possibility of fraud or not”, he says.
He adds: “So with this, what we are achieving is stopping a large number of frauds, detecting them in real time. It’s having the police at the front door, so to speak, instead of coming in and doing forensic surveillance of what happened after the fraud has been executed.”
Another example he mentions is if at a point in the Mexico City a card transaction is made, and two or three minutes later a new one is attempted but on the other side of the capital. “So I have the information that comes from the point of sale terminal at that moment, and I can determine its geographic location and say: here is something that doesn’t make sense to me,” he says.
More business opportunity
But this new technology in the processors not only serves to prevent fraud in real time, but also to open up the opportunity to offer more products to users, according to their needs at that time, using artificial intelligence. .
“What is being done is to contextually analyze the characteristics of this transaction, and it is not just doing it for fraud, but it is also doing it to offer a new income opportunity for the bank,” he emphasizes.
rrg