Whenever somebody wins an award and they have to talk about it on TV, they always say the same tired thing: “I am humbled to receive this.”
C’mon, that’s garbage. As soon as they get the call from their agent, or their mom who sees it online first, they jump up and down and grab a beer and call their girlfriend and even the neighbors can hear them.
It’s not as much fun jumping up and down here because we’re mostly remote (there’s that virus thing you might have heard about), but we’re definitely happy to learn that we’ve won a coveted Finovate award. Socure received this recognition specifically for “Best use of AI/ML.”
There are a lot of great vendors who got Finovates this year, but we’re happy to get the one we got. Because artificial intelligence and machine learning are the foundation of our company’s solutions. Although those are in the wrong order. Let me explain.
Machine learning builds your curriculum. Artificial intelligence then delivers it. ML tells the Tesla what a road and traffic looks like. AI drives the car. So what’s Socure’s version?
A lot of other vendors who claim they employ machine learning really mean that they build lists, maybe some rules. “Here’s our pile of fraudulent emails. We better not find you in there, or you get rejected.” Yeah, well, anybody with a clipboard and a legal pad can do that.
Socure leverages ML to ingest, cleanse, and learn from hundreds of millions of outcomes, transactions both good and bad, in order to create models for validating identity. We learn the patterns of honest and dishonest applicants. Sure, we build lists, but we also learn how to react to somebody we haven’t seen before, or maybe we have, but the last ten times they applied, they used a modified set of identity elements. If we don’t know what you look like, we know what you’re supposed to look like.
True machine learning means processing huge volumes of useful data and deducing from it how to make decisions. And we layer these lessons on top of deep feature engineering. Because we also learn from huge tracts of raw data what a good/bad email, phone, address, person look like. Do I like all the fields you submitted on your application? Maybe. Then do I like them all as a combination? Just because those items are all good doesn’t mean they all belong to you, since you might have stolen them, maybe even from multiple victims.
Okay, so there’s ML. Now the A.I. kicks in. Our customers get applications from their own customers, consumers who want credit cards, loans, account openings. They submit those apps to us. Our ML-based models kick in, and the A.I. helps generate the decisions, the scores, the reasons for those scores, and returns those to our clients who can then decide which apps they like, and which ones they don’t.
We’ve also automated this whole complex process. The objective is to make our platform smarter and smarter, faster and faster. Instead of a whole bunch of people sitting in a room saying, “Hmm, how do we make our solution better,” we let the data tell us that, in high volume and at great speed. We’re better than 97 percent accurate in distinguishing good applicants from bad, then providing our reasons for our decisions to our customers so they can decide how to react.
Not to say that we don’t need the people. A large percentage of our staff is comprised of data scientists who are smarter than the rest of us, who do all the other stuff that needs to get done.
So when we win a thing like the Finovate award from that great AI/ML, the DS staff gets to jump up and down, while pretending to be humble about it. But I’m important, too. I mean, somebody had to write about it.