From Guardrails to Growth: Unlocking User Intent with AI

0:00
So uh thank you all for joining uh today’s session. Um it is early on my
0:05
time so filled with coffee if I speak too fast. Uh hopefully Simon you will
0:10
let me know. Um today’s session as Simon mentioned is um from guard rails for
0:15
growth unlocking user intent with AI. Um if you haven’t heard uh I do work with a
0:22
company called SIFT. It is a AI fraud decisioning solution for uh companies
0:28
like Alaska Airlines, Adam Tickets and Tap Tap Send. And as Simon mentioned, I
0:34
spend the majority of my time at the intersection of product road maps and
0:39
really getting into the nitty-gritty of our customer challenges really to ensure
0:44
that the innovation that we build in our labs is practical in the real world.
0:50
What I’d like to start with is um the concept of what is user intent. You
0:57
know, uh with every payment method and touch point and demand for more
1:02
convenience, um I think a lot of these operations in
1:06
a sense are trying to operate blindly and they’re trying to understand in in a
1:13
digital environment, can I essentially trust this person? Um do they mean me
1:18
harm? Will they be profitable? What is their intent? And really understanding a
1:24
window into the why I think is the future of fraud prevention. And I hope
1:29
by the end of this short session, my goal is really to show you how AI and
1:34
identity intelligence can really help you understand what is going on your
1:38
platform, but really the why behind it. So to start, I’d like to share some
1:45
interesting stats that our team of experts have collected over this year so
1:50
you can understand um what the new fraudster kind of looks like. Um in a
1:55
study done earlier this year, we found that 34% of consumers have seen offers
2:01
to participate in fraud. You may want be one of those 34%. Uh 35% have personally
2:08
participated in fraud. And if you are not even in that 35%,
2:13
65% um have personally participated or or
2:17
known someone who’s personally participated in fraud. And so what we’re
2:22
seeing is really just this democ democratization of fraud. I can never
2:27
say that uh word right. Uh but fraud has essentially kind of gone viral. Um and
2:34
so now what we’re also seeing is the reason why um the access and the howto’s
2:42
are becoming easier than ever. What you see on your screen from left to right is
2:48
really real dark web research on how accounts are being sold, how credit card
2:54
information or gift cards are being sold, um, and how new technology is mim
2:59
mimicking browser activity. And most interesting of all really are fraud
3:05
guides. You see now kids using things like Tik Tok um to show best practices.
3:13
And so this is my thought that I try to leave with everyone is that becoming a
3:19
fraudster has become much easier than it’s ever been. And fraud being a
3:24
fraudster has become something that is more essential to being a side gig.
3:31
And a key reason why is because when you look at the demographic of the payment
3:39
users nowadays, they’re just getting younger. We see in the last five years a
3:44
two to 3x times in the type of digital wallets that are being produced. And
3:48
what we’re also seeing is a skew in the to risk, right? And so what we’re seeing
4:08
when you zoom out essentially is a uh spread and scale and lowering of the bar
4:15
of being a fraudster and also payment methods that are drawing in younger
4:20
digitally native users. And so the wave of the challenges that are coming
4:27
towards businesses who do any type of payment is not something I envy in terms
4:32
of having to change. And something that was mentioned earlier is that the fraud
4:37
use cases that that tie to payment are expanding. Policy abuse for example is
4:42
becoming something that is a bleed that has become quite fatal. And so the
4:48
question then becomes, you know, are they getting more
4:53
sophisticated? Right? Um it sounds scary. Um, it is scary the speed and the
4:58
scale and you’re looking at uh an environment where you’re like, can I
5:04
outsmart them? And the answer is kind of yes and no. When you look at how
5:12
fraudsters behave, here are some user attributes that are usually used. And
5:17
for the most part, being a fraudster is still a numbers game. Buying info on the
5:23
dark web is quite easy. What is hard however is the reuse of identity.
5:30
Earlier they mentioned uh you know synthetic ids and automations and making
5:35
front door uh technology less and less infective like KYC and it’s because
5:40
they’re they’re mimicking a lot of the things to create the digital footprint.
5:45
Um, and so what we’re kind of seeing, sorry my browser is a little worried, is
5:51
that um, despite all of the tools and the automation that is happening to
5:57
create a digital identity, it’s still quite hard. Fraudsters really are just
6:03
about volume. So, if they take a moment to go through and create a synthetic ID,
6:10
you can be guaranteed that that investment is something that they’re
6:14
going to continue to use across different platforms. And so, I’m going
6:20
to go through a really quick uh scenario that is quite common. A fraudster buys
6:25
stolen personal details. Not hard. maybe one person’s name, another person’s
6:29
address. They stitch them together into a new fake person, a synthetic ID, a
6:34
digital person. They’re using this fake ID to sign up for a mobile wallet,
6:39
passing through perhaps quick onboarding checks because that business perhaps is
6:43
trying to grow. They don’t want to apply a lot of friction. They link stolen
6:47
credit cards or take advantage of sign up uh bonuses and referral rewards and
6:52
they m they move money between multiple fake wallets accounts or cash out really
6:58
before anyone notices and then they really move on to the next company. The
7:02
truth is uh or the reality is is that they will come in fast, they will hit
7:07
you hard and they will leave. And a lot of businesses can be left blindsided
7:11
because to to a lot of these businesses, this user looks pretty legitimate and
7:17
they look like a new person. And so this goes to a lot of what was spoken a lot
7:23
uh earlier about the fireside chat is that you know um a lot of the guard
7:27
rails that are put in place are becoming less and less effective. And at this
7:32
point in time, it’s really important to look I think what you said from the
7:36
outside in. If you look at the scenario to that
7:40
business, that user looks new and that user looks pretty legitimate and the
7:43
user moves fast enough for you might to might not be able to track it. This is
7:48
where things like AI and identity intelligence and a global network really
7:53
become key. So the question that I’d like to ask is this,
8:00
what if you knew? What if you knew how long that user’s digital attribute has
8:06
been used around the globe? What if you knew it was linked to a history of past
8:12
risky behavior or trusted activity on the flip side? What if you knew in
8:17
detail which industries when and where they were acting? And what if you knew
8:22
even more, I think poignantly, a detailed peer review of other fraud and
8:27
payment analysts looking at the behavior prior to hitting your platform.
8:33
This, I think, is what really drives this concept of context. Context is a
8:39
holistic look at the user and with this additional type of context, you can get
8:46
intent. We are trying to mimic the natural logic that we have when we meet
8:52
someone new. The more people we know that know them, the more we know about
8:56
their behaviors in different environments, the more we know about
9:00
their past, the more we have an idea about whether or not they intend to harm
9:06
us. What level of trustworthiness do they have?
9:11
And so this is what you know I nerd out to be honest on on this concept of built
9:17
in uh context. Uh we recently launched something called identity trust XD. I
9:24
can go into a lot of technical terms but to boil it down it looks at a user in a
9:31
new light. We can provide behavior beyond your business, behavior breaching
9:37
platform norms and behavior um breaching user patterns. This is the combination
9:43
you can combination of AI and um uh global network information and identity
9:50
intelligence to really give you a holistic look at the user looking I
9:55
think what he said outward in. But then when you pause and look at what’s in,
10:00
another thing that I find a lot is that despite how quickly and how cohesively
10:08
fodsters are, departments inside are pretty siloed still between the
10:14
compliance and the payments and the fraud team and the product team and the
10:18
customers pushing promotions. um they’re all moving in somewhat of the
10:23
same direction, but what is lacking is a cohesive look at the user themselves.
10:29
And so what we see here is just a quick snapshot, and I like to show this to to
10:34
some folks, is really journey mapping, but journey mapping inside internally,
10:38
right? What you’re really looking at is user signal. The same user at the same
10:42
point of the customer journey, but different departments looking at it from
10:46
different views, but they’re looking sometimes at the same signals. And so
10:49
what will happen in practicality is that someone will launch a new payment
10:53
method. That payment method actually has different behavior, a digital wallet,
10:57
let’s say, than a than a um real time a different type of um a but that insight
11:05
in terms of the movements move a little bit too slow across these departments to
11:10
be effective. And in the fireside chat, speed is absolutely key, right? And so a
11:16
different way to look at intent really is the speed and the collaboration of
11:21
the organization to see the user in real time in totality across these different
11:26
departments to really understand a holistic uh look at intent. And so at
11:32
SIFT, we’re kind of taking insight to a whole new level to really
11:40
understand not just behavior, but really the intent behind it. And I and I truly
11:46
see that this is something that has been uh game changer is such a tech word, but
11:53
I will use words that our customers use. Decision and confidence, speed, trust to
11:58
acceleration. And so um you know uh if you visit our website you’ll hear um
12:04
from a fraud analyst who was one of my favorite people to work with uh in
12:08
developing the solution um speak about how this type of layered insight
12:14
crossmental insight has really changed the way that they look at the user. And
12:19
so, uh, I really thank you for your time today and I hope that this sparks a
12:24
little bit of a conversation whether it be internal silos or, um, you know, how
12:30
you look at your business and look at your user moving forward as you are
12:34
adopting new payment methods.
12:37
>> Thank you so much, Stephanie. Really insightful um, discussion there and
12:41
definitely uh, something that’s going to raise a lot of questions uh, which I’m
12:44
sure anyone has. Make sure you do use the Q&A. Uh with that first question
12:49
then is how can businesses balance fraud prevention with seamless intent driven
12:54
user experiences.
12:55
>> Ah that is such a great question. The key is uh million-dollar question right?
13:02
Um I will say um
13:07
automation insight and speed three other big pillar
13:12
words right uh but I will we’ll go back to the fireside chat. It’s really
13:16
testing. You really have to have the unique insight um
13:22
across the the different departments. I like to think of them as Lego pieces.
13:27
Lego pieces are behavior attributes um that can be pitched together to
13:33
understand what a behavior should and should not look like at each point of
13:37
the customer journey. So a lot of people frontload the budget in like KYC and
13:44
ensuring the front door, but the front door is something that is becoming less
13:48
and less effect harder and harder to protect. And so you really have to look
13:52
at the behavior of the user across the customer journey. And by being able to
13:58
essentially play around in refinement at each part of the
14:03
customer journey, you can adjust friction more specifically. And uh we
14:09
have something called uh workflow simulation. Essentially, it it allows
14:13
you to simulate a lot of these changes in real time. So let’s say before a big
14:17
launch of a product or launching into a new segment, you may want to play around
14:22
with that friction and see what that looks like in a safe environment.
14:26
um and then uh apply that. Another thing about the cross departmental uh uh
14:32
collaboration is that something might happen in the security team or the
14:37
payment team and um what we see is a lot of reactionary responses. Let’s block
14:42
them, right? Let’s block this IP address. There’s pretty big sledgehammer
14:47
responses. um the ability to implement little Lego pieces and details and
14:52
insight from different departments quite quickly into the automation and test app
14:57
prior to we really help kind of that balance between growth and security and
15:03
um and uh um fraud.
15:07
>> Thank you so much indeed for the question and for answering that one as
15:09
well. And also looking at AI uh can it help uncover user intent to drive both
15:14
trust and growth? Stephanie?
15:17
>> Yes. Um, I think the pre previous speaker kind of mentioned a little bit
15:21
about, you know, there’s so many different ways to use AI and AI, I live
15:27
in Silicon Valley. You cannot drive down San Francisco without seeing some
15:31
billboard of some AI that’s going to help your life. And all of it just
15:35
sounds like buzzwords, right? Um, in a in a lot of ways, it is a big buzzword,
15:39
but it it really truly is quite powerful if you use it correctly. M
15:47
>> and um with SIFT we take kind of a um a very powerful but pragmatic
15:54
approach uh where the AI is married quite closely with the human user
16:03
review. your team, the fraud team, the payment team, the security team, heck,
16:08
the customer team, the customer service team knows more about the business,
16:12
which and how that moves, which is reflected in how effect how effective
16:17
the AI can be. And so, uh, there’s different models, but we use a super
16:23
supervised learning machine, which means that, um, a lot of the analysts do
16:28
tweak, uh, the models themselves. That sounds like sounds very difficult but
16:32
it’s actually quite not. Um and in doing so the model is trained specifically
16:39
just for um that the that model of business. What is also unique about our
16:45
approach is that a lot of the times you see AI models being built um or run on
16:51
them on their own, right? Um, but as in the fireside chat, fraudsters payment
16:58
methods are more global than ever. And being able to embed that uh global
17:04
insight and even that industry insight into the model is what’s really really
17:09
key here. And with Syth, we kind of take a multi-layered fortified approach where
17:15
we have a global model to track those payment trends uh across our network, an
17:21
industry model, which is quite it we I think it was last year, but it was it’s
17:26
actually quite powerful. It takes specific industries um and builds in
17:31
behaviors between the different industries and layers that on top of the
17:35
custom models to really give you that kind of inside out and outside in view
17:39
of those models. And those can really drive a more intelligent automated
17:44
friction um solution.