Responsible Gambling Reinvented: Protect Players and Margins Together

0:03
Hi everyone and welcome back to the final session of today’s SBC digital
0:11
player protection day. Um it’s been a fascinating day and it’s going to
0:16
continue into our final session because I’ve got another great panel um where
0:21
we’re talking about uh a very important issue um protecting players and margins
0:27
at the same time under the uh nice title of responsible gambling reinvented.
0:35
Um so um my name is Steve Hall. I am the
0:40
editor of the player protection hub at playerp protectionhub.com. Go and check
0:44
it out if you don’t know it already. You probably will do if you’re here. Um I am
0:50
here with four great panelists. We have Andrea Carvalia, the head of legal
0:56
compliance and MLRO at NA. We have uh Alexander Hall who is the
1:04
trust and safety architect at SIFT. Um, all the way from Las Vegas. Uh, Andrea
1:11
joining us from is it Lisbon, Andrea?
1:14
>> Yes.
1:14
>> Yes. Yes. Also from Lisbon. Yes. Yeah. Miguel Luis, the head of compliance at
1:20
Portuguese operator Labul. Uh, and last but not least, she got here
1:25
just in time. It’s Stephanie Trin, the senior product marketing manager at
1:30
Syft. So, welcome to you all. Thanks. Thanks very much for joining us.
1:36
This session explores how leading operators are bridging the gap between
1:40
fraud pre prevention and responsible gambling using AIdriven identity and
1:46
behavioral intelligence to detect risk early before it becomes a regulatory
1:52
issue, a brand problem or a drain on net gaming revenue. So, um, first of all,
2:00
Stephanie, I’d like you to kind of explain how you view this overlap
2:07
between player protection, fraud, um, and and payments as well, I guess.
2:13
>> Yeah. Um, all I can say is I do not envy the the responsibilities role of both
2:19
Andrea and and Miguel here. It is a tough situation to be in and fast
2:23
changing even for the pace of I gaming. um the way I see them merging uh we used
2:29
a lot of big words around identity and and you know AI but from a 30,000 foot
2:35
level in speaking to operators and uh folks in the industry what we’re see
2:40
what we’re starting to see is a real mergence between financial crime risk
2:45
and player harm risks. These are traditionally two separate initiatives
2:51
with some overlap, but they’re starting to merge more and more so proliferated
2:55
by things like irresponsible gambling initiatives. Um, what they’re both
3:00
rooted in, however, is in understanding very nuance player behaviors. And these
3:07
behaviors are getting more complex with things like promo abuse, new launches,
3:12
new regulations, um, and new, um, game types. And what I
3:20
see is that as responsible gambling becomes uh, a bigger and bigger topic,
3:26
all of those things fall on folks like Andrea and Miguel. Um, and I’m going to
3:31
walk you through kind of a a story um that showcases really how they kind of
3:36
overlap. So, let’s say I am a player. I’ve self-excluded myself. I’ve had a
3:42
moment. I don’t I don’t want to, you know, gamble anymore. And then I see a
3:47
promotion. The promotion is so enticing. I can’t help it. Uh, I know that I’ve
3:52
self-excluded myself here in this one game, but I’m going to go and try use
3:58
maybe a little bit of a different credentials, maybe a different email
4:02
address, a different payment method, and I’ve now been able to create a new
4:06
account in a different game. Using that new account, um, I’ve added new payment
4:13
methods. I’ve used promos really quickly. I started to do binge wagering,
4:17
meaning that I’m making a lot of wagering at a, you know, overnight so I
4:22
don’t get caught. I’m making rapid withdrawals. Um, all of these behaviors
4:27
actually occur after the point of login. Let’s say at certain points KYC fails
4:33
because there isn’t a automatic universal KYC that once you get
4:38
registered for self-exemption, it gets passed around everyone, right? And so,
4:44
uh, these are examples of behaviors of that player, in this case me, that
4:51
actually span both payment, fraud, CX, and now falls into the lapse of
4:57
compliance. These are behaviors actually that are uh pulled from recent um recent
5:04
lawsuits. I won’t name names, but this is a compilation of behavior signals,
5:09
none of it on this call, I believe, that have had real regulatory uh um
5:14
consequences. And part of the takeaway from that is that there were gaps in the
5:21
ability to see those behavior signals because they stereotypically fall in
5:25
different departments around different silos and they may seem okay at certain
5:31
points. Let’s say the fraud department sees their specific lens is okay, the
5:35
payment is okay, but it’s actually the totality of how that player is able to
5:39
move or is moving through all three departments um that really give the um
5:45
the the spikes in risk um that that happen. And so this is where we kind of
5:51
see these traditional risk signals that are very much focused have been focused
5:55
on more financial crime starting to become more and more relevant as we look
6:01
at player harm risk.
6:02
>> Yeah. Stephanie, if there’s lawsuits around these, and I’m going to slightly
6:08
put you on the spot. Surely they’re in the public domain. So surely you can
6:11
tell us about them.
6:13
>> Yeah. Well, I will I will let it I will I will leave it anonymized and feel free
6:18
to research. I think the the takeaway though is that you would think some of
6:23
the behaviors that I just went through were very very simple, right? Maybe and
6:27
there are learnings as they come along. But um like I said in the beginning, the
6:33
landscape of how players operate now are changing so rapidly by force function of
6:39
promos, by force function of new games that are put in place that have semi-
6:45
different velocities based on the game types. These lessons learned um are
6:50
really fast and and not relevant for very very long. And so while they seem
6:56
simple to detect, they’re actually rooted in something that folks like
7:00
Andrea and Miguel may not have um uh control over, which is it’s really
7:06
faulted in a deep silo between the departments both operationally and how
7:12
they communicate and also a lot of the technical integrations that that are
7:18
required are quite quite deep. And so if you can and you Google and you dig in,
7:22
you’ll kind of see where those data uh integrations kind of fail. So very
7:26
techy, hence why we started with words like AI and identity. But at the end of
7:31
the day, it’s almost often basic behaviors, but when you look
7:38
at at the scale of, you know, what I gaming operators come through in terms
7:42
of volume, it’s kind of difficult to detect the nuance. And like you said in
7:46
the title, trying to balance those controls and not, you know, lock players
7:53
too early um or too frequent is is definitely a challenge.
7:58
>> Yeah. Um you talked about silos there. Let’s let’s talk someone who doesn’t I
8:03
don’t think she’s got any silos in her organization because look at the look at
8:08
the size of her job title. Andrea, you you your job title uh head of legal
8:15
compliance and MLRO suggests quite an overlap certainly in in your
8:20
organization between legal compliance and AML. Can you can you talk to us a
8:27
little bit about your job and the structure of the organization which
8:31
which which might feed into this topic?
8:35
>> Thank you Steve. It’s a pleasure to be here. Um and uh my background is in law
8:40
uh uh but my daily work is really about you know connecting three worlds uh
8:44
regulation uh player safety and and business performance. So this overlap of
8:50
course there is a strong overlap between AML and responsible gaming. um
8:54
behavioral patterns that might suggest money laundering as uh as Stephanie well
8:59
said like large deposits by quick withdrawals or similar examples that
9:04
Stephanie indicated can also be early signs of gambling arm. So when we
9:09
provide the source of funds at we are not just uh preventing financial crime
9:16
we are making sure that the play is sustainable. So both AML uh and
9:21
responsible gaming rely on data on continuous monitoring and on human
9:26
judgment. So the same data that help us detect money laundering also helps
9:32
detect ARM.
9:34
>> Yeah.
9:34
>> So yeah so this is really important to have all all of these areas connected
9:38
all the systems all of the data. Um from maintain um we we operate in many
9:44
jurisdictions. Uh so consistency is key but so is local relevance. We use a
9:50
global um compliance framework but each market has local ownership on the
9:56
responsible gaming and AML procedures. For example in Portugal source of funds
10:01
um and AML reviews are aligned with the local income data
10:06
while uh UK in German or even European markets similar triggers are higher but
10:13
more granular. So consistency builds trust but localizations build relevance.
10:20
And see if you mention a really good point regarding the silos point. Um uh
10:24
yes this is a real from our experience is a really good um point of discussions
10:29
among us. Um because this is one of the biggest challenges for for many operator
10:35
and also Stephanie indicate we have to work with many areas different data
10:39
different systems. So at in we have built uh crossf functional uh
10:46
governance. So we have committees um we have many discussions that brings
10:50
together all AML responsible gambling uh fraud product and business etc different
10:55
areas all together to discuss we share dashboards alerts KPIs so everyone works
11:02
from the same data but even more important than um the instructure is the
11:08
culture so we have worked hard to make sure that compliance uh is a partner to
11:13
the business not a blocker So our our aim here is replacing silos with share
11:19
accountability.
11:21
>> Yeah.
11:21
>> Um and from from your last points to regarding my role, yes, my role brings
11:27
together uh legal compliance in ML. Uh and that is it is by design. We don’t
11:32
see them as separate boxes. Uh they are all part of a risk ecosystem from the
11:38
company. So in in some in summary, our mission is make sure that every control
11:44
from AML checks to responsible gaming tools protects not only just the
11:49
business but the player. So structure is consistency across Sentine. Um local
11:55
teams have to adapt their market to to the global frameworks but the philosophy
12:00
is the same. Uh we manage risk holistically not in silos.
12:05
>> Yeah. It’s interesting as well because if you look at um regulatory actions, I
12:10
mean Stephanie mentioned lawsuits, but regulatory actions certainly in the UK,
12:15
the the reg the gambling commission will always site AML failures as well as
12:21
responsible gambling failures in in the same breath. So
12:25
it’s not it’s not just um it’s not just you that thinks of it as as the same um
12:32
the same ballpark. Um, Miguel, you you talked very eloquently in our prep call
12:38
about balancing the need to make um profit with with compliance.
12:43
How do you achieve that balance? And can you give us some real life examples
12:49
perhaps where profit might have compromised compliance or or hopefully
12:53
more the other way round?
12:57
>> Yeah. Uh thank you, Steve. So as Andrea and Stephanie were already hinting, this
13:02
has to be a joint effort. It cannot be siloed. It’s not going to go well on on
13:08
that perspective and with those uh methods. It has to be a joint effort
13:12
between all the relevant teams that interact with this point. Uh because the
13:18
let’s call it traditional stereotypical view of compliance is a bit of a
13:22
business blocker which it it isn’t. If at most right now we’re trying to the
13:29
most possible to be a business enabler or at least to help businesses
13:35
flow uh and navigate in a safe manner that is uh good for the business but
13:41
good for the players as well because uh at the end of the day it has to be good
13:46
for everyone otherwise it’s good for no one in the long term at least. Um right
13:52
here from from our perspective we can do it uh on a thing that seems kind of
13:57
basic if you talk about it but actually makes makes a ton of sense. This is
14:02
checking how the player behaves because again you the team here was speaking
14:07
about um different jurisdictions that okay you have to keep to build strong um
14:14
frameworks globally but then adapted to the localization because there’s always
14:18
specific constraints on each jurisdiction. Uh I think we can go a
14:22
little bit bit level deeper on that. We have to adapt it to the player itself
14:27
because different players have different risk thresholds. Different players have
14:31
different activity patterns. Uh what a thing that could be a risk indicator for
14:37
a person might just be normal activity for another one. I I I keep seeing some
14:43
colleagues uh when they see a player doing deposits on odd values like
14:49
certain not 100 but €13742. that on the perspective of us, everyone
14:58
else already laughing because we both know
15:02
that’s kind of fishy. But for some people, that’s just normal. They’re
15:06
they’re lucky numbers or they did some calculations. That’s how much they need
15:11
to do something in specific. We have to tell our things not on just a
15:17
global level, then on a specific jurisdiction level, and sometimes even
15:21
on a specific player level because it can go to that level.
15:26
But keeping this healthy for everyone, it’s it’s very important because without
15:31
the players having a good time and enjoying gambling as a entertainment,
15:35
which it is what it should be, uh it’s never going to be good for anyone at the
15:41
at the end of the day, not even to the business, which is what we’re speaking
15:46
here. Yeah, I
15:47
>> I really love these these themes and uh they sound simple, you know, uh and I
15:54
but however, I can see the smiles between the two, they are they’re giant
15:58
organizational shifts and kudos to you two for for really driving that that
16:02
effort. Um the the concept of sustainable growth,
16:07
shared accountability, I absolutely love and you know uh having to standardize
16:12
yet personalize. I mean they’re they’re opposite ends on on each side of of
16:17
those conversations. Um and you know very much focused on the application to
16:24
IAEM but extracting it out in terms of the the the title itself about
16:29
profitability. Um the the root of it is very technical. It’s data and controls
16:36
and automations. Um but it is rooted very much in a in like what Miguel had
16:42
hinted a very very deep and nuanced understanding of the player and I can
16:46
think of it uh as the digitiz digit I can never say that right digitization
16:52
let’s say e-commerce okay you know e-commerce opened the door for a extreme
16:59
wealth of data about the consumer in a digital platform in one hand it can be
17:04
very overwhelming but that data essentially over time has been used to
17:09
create very targeted person uh personal person personalizations as you guys have
17:13
seen in your own personal ads that has shifted into how products are developed
17:20
how uh the journey is um personalized for each type of consumer and each type
17:26
of persona and so while that has happened at a pretty high clip rate for
17:31
things like e-commerce you know i gaming the foundations of that is is really
17:35
starting I think to to take root and folks like Anderia and Miguel, you know,
17:40
um compliance is always kind of seen as a blocker, but uh to both your points,
17:46
it is a catalyst for that type of deeper player understanding in a digital
17:52
environment and personalization.
17:54
>> Yeah. Um you talk about sustainable growth there. Andrea, I’m just going to
17:58
turn back to you just because I feel like Entain was one of the first
18:02
organizations that kind of talked
18:07
about s sustainable growth. Um would you say it’s been on a how how I guess how
18:14
would you describe that journey and and kind of have you noticed um a cultural
18:21
change um at the organization?
18:25
>> Yes. Yes. This is really important for us. We see uh sustainability as a a
18:30
really good point and uh and as a focus for for all of the all of the market uh
18:38
growth and and company growth. So we consider that um responsible gambling
18:44
and complying are not just about regulation, it’s about trust. So when we
18:49
have when we have trust in place uh and sustainability, it protects both the
18:54
player and the business. So if I had to summarize our philosophy in one line, so
19:00
compliance and player protection are like two sides of the same coin. So
19:04
sustainable engagement. Um if if we look like five years ahead, uh I believe that
19:11
the future of compliance and player protection uh will depend on integration
19:17
and empathy. uh integration because uh the data must flow across teams and we
19:23
can see that data will be our future AI and we have to absorb that also in in
19:28
organizations and use it responsibly. Uh but empathy because of in the end of uh
19:34
each transaction um there is a person uh not a risk score. So technology will
19:41
help us to see the whole player. Uh also like Mikuel indicated and and Stephanie
19:47
will indicated because we have to see u also the compliance uh review the player
19:52
review. Uh but also technology will help us to see everything the whole player
19:56
behavior but human judgment uh is will that the key to keep the industry safe.
20:04
>> Yeah. Yeah. Interesting. Alex, I’m just going to turn to you. um in in in the US
20:10
um after years of investment um over the last probably two years I guess
20:16
investors and analysts have started to demand EBIT DAR positive results that
20:21
was that was the phrase I would hear in every quarterly report when are you
20:25
going to be EBIT DAR positive when are you going to be oh soon Q3 Q4 anyway um
20:32
>> h has this meant a detrimental effect on player protection policy
20:38
in the US do you think
20:40
>> I am sure that it exists somewhere to say to be yes I’ll be fully transparent
20:46
somewhere behind some closed doors the answer is going to be yes however with
20:50
everybody that we work with the storyline extends back from i gaming
20:54
into the other industries like Stephanie was mentioning where uh fraud has grown
20:59
to a place beyond payment and into account focused fraud where these
21:05
intangible metrics become very important such as brand reputation, customer
21:09
trust, customer loyalty. These things that are very hard to quantify have
21:13
become commonplace in our KPIs and our metrics in some way, shape or form. Now
21:20
bring that forward to i gaming and we realize that it’s it’s a continuation of
21:25
the same storyline where these uh unquantifiable inquantifiable uh metrics
21:31
become very important and through that player protections become important
21:35
because brand reputation uh in these highly competitive markets brand
21:39
reputation has become very important. So when it comes to the those those fights
21:44
between Ebida and uh and brand reputation and through that player
21:48
protection uh by and large player protection is is definitely respected.
21:53
>> Yeah. Um and it’s all different parts of the the same ballpark, isn’t it? Um
21:58
Andreas, what what um sorry Andrea added an s there accidentally. I’m
22:07
sorry. I know that makes you a madman. Anyway, what what guardrails do you put
22:12
in in place um to to protect uh yourself?
22:17
>> Uh that is a really good question, Steve. Um in our experience um we use a
22:23
layered approach to guard rails. Um system uh people and processes. So
22:31
system are automated for example like players limits uh self-exclusion uh
22:37
transaction monitoring um people are are you know everyone that’s go through the
22:42
training um and have clear accountability on on what to do on these
22:46
kind of situations and um and processes are about continuous review. Um the
22:52
businesses have to continuously evolve and every control should be effective
22:56
but not excessive that can block the the business itself. So we actually track
23:02
how these measures impact profits uh not to reduce them but to prove they sustain
23:08
it. So when we protect players uh you protect the brands that and that
23:15
protects long-term profit and that has an impact on on on the on Alexander
23:20
comments. So if in one line compliance again for all the audience doesn’t kill
23:27
profit, it protects it.
23:30
>> Um we are also seeing and and also recing again from from from Stephanie’s
23:34
comments, we also seeing a lot of new fraud typologies, synthetics IDs, uh
23:39
bots, arbitrage, collusion and even accounts mules uh that are being used by
23:45
pass checks. So these cases show how fraud AML and player protection are
23:53
interconnected. Um you can’t treat them separately anymore. So the risk here is
23:59
the blind spots between all of these areas. That is a real risk. But that’s
24:03
why you need to work together.
24:06
>> We have a book on Stephanie.
24:09
>> Oh uh we have a saying I think it came from Alex but fraud exists and thrives
24:14
in fragments. And Andrea, you just made me think of that.
24:20
>> Yes, it’s a really good point. Yeah,
24:23
>> great areas and worse.
24:27
>> Yeah, just think there’s there’s an additional point here that could be
24:32
made, which is we’re talking a lot about data and it’s crucial data. It’s what
24:36
allows you to validate the information that you have or that you don’t have.
24:41
But one key aspect that we tend to underestimate sometimes is the
24:48
subjective very subjective evaluation of the communications of the player. Uh
24:54
speaking with the customer support team, speaking with the teams that directly
24:59
deal with the player, checking the player behavior in terms of responses,
25:04
the responses that they give, the the way they approach the company through
25:09
customer support. Usually uh the the tone the the contents of the message
25:15
that usually tells a lot of story behind the data that can be
25:21
of great use use usefulness to to assess the healthiness of that relation
25:28
business relationship between the the company and the player. And even in
25:32
these cases like we were speaking about fraud because usually through through
25:36
the communications we can get a lot of underlying information that can help us
25:42
uh decide if a certain player or a certain transaction is to be trusted or
25:48
not. And I think that we tend to undervalue the that very subjective
25:55
again uh key aspect of the the communication. But we should take it all
25:59
very in consideration because it can help us uh guide the same way compliance
26:03
does help guide the business where it should go and where the the
26:08
communication with that player and the interactions with that player should go.
26:13
>> Yeah. Yeah. Interesting. Um just at this point I’m just going to have a slight
26:17
interruption to bring attention for the audience to the uh buttons on the right
26:24
of your screen. Um in the middle there you’ve got a Q&A button. If you press on
26:29
that then at the bottom it says ask a question. You can click on that and and
26:33
send us any questions you want and I’ll try and get get them answered. Um so um
26:40
Alex, I’m just going to go back to you. Andrea talked a lot there about the
26:45
sophistication of fraudsters, about bots and so on. Um, we haven’t mentioned AI
26:51
much yet. Do do we use bots to fight the bots or
26:57
um tell us a bit about the technology?
27:00
>> Uh, we absolutely do. So, um, I would like to encapsulate what Andrea and
27:06
Miguel both said. Uh, Miguel brought up social engineering really through
27:10
elements of social interactions and these subjective interactions and how
27:13
can I get through the platform and then Andrea brought up the the idea of
27:17
synthetic identities um, identity theft, mule accounts and all of these different
27:22
things. And Stephanie had brought up the idea of behaviors and the idea of
27:26
relationships between maybe first party and second party in order to get to get
27:31
past these player protection checks and compliance checks and all these
27:34
different things. So when we think about all of this these different
27:37
methodologies these different ways to enter and transact uh abuse or exploit a
27:41
platform all these different things uh you can imagine there’s a massive amount
27:46
of data associated with it all of these different data sets come into play all
27:49
of these secondary and tertiary implied behaviors and storylines come from it
27:54
and so yes when it comes to the data as SIFT has done for the past 14 years AI
27:59
machine learning is pivotal in order to process all of that and extract the
28:03
relevant storylines Now, one thing that hasn’t really been
28:07
discussed here so far is actually uh the the reason why SIFT is is pushing into
28:12
this this this narrative of player protection is how we resolve
28:18
information across a network instead of just res uh resolving around a single
28:24
identity and who this person might be, who they might be transacting as as
28:28
whether or not we should trust this particular uh account or transaction.
28:33
Instead of that, uh, instead of only that, I should say, we’re going up to
28:37
the entire network and extracting all of these behaviors across all of these
28:41
different player platforms and bringing that to anybody that chooses to work
28:45
with us, of course. And all of that is prior to Agentic AI, right? So now as
28:52
we’re heading into 2026, the world of possibilities that falls
28:57
under the umbrella of Agentic AI is so expansive that it’s hard to put a button
29:02
on it, right? But uh we are definitely looking into the ways that uh automation
29:09
can be handled via AI instead of just simply aggregating data which is is no
29:14
simple task. How can we automate using you know AI in different isolated areas?
29:21
Um we’re promoting that to different partnerships and different uh companies
29:25
that we work with and yeah just the world is is insane as we
29:31
head into to bigger uh bigger world with AI agentic AI. But as far as if you want
29:36
to get into specifics about the AI and what you know our models consist of
29:40
Stephanie is definitely the one to ping for that.
29:43
>> You know um from a from a technology perspective uh there there are so many
29:47
different that is my dog. I apologize. Uh several different layers. Um one
29:56
second. Um
30:00
>> um
30:00
>> if if you want if you want to take a break to put the dog out the back door,
30:06
we can come back to you
30:08
>> back in. Uh what I do want to go back to though is this concept that uh both
30:13
Andrea and Mel do and why AI is is is incredibly important. Um what what
30:19
Andrea talked about is uh integration and empathy and this is where AI can
30:26
really take large large data sets and trying to find those nuances in those
30:32
behaviors while allowing um still a very human touch to the approach. And so
30:39
there’s many many different ways to uh look at AI and use AI, but really the
30:45
takeaway is the core of what Andrea said is about the integration piece is being
30:52
able uh to have a centralized point uh to ingest the data from different
30:57
departments, analyze that data in a much more automated way and to leverage the
31:02
expertise of folks um like Miguel said to really understand the player. And
31:09
what AI allows uh folks to do is not only see a more holistic player insight
31:15
uh player journey, but uh to forecast, you know, how to kind of put in
31:21
regulations or policies going all the way back to the top about profit because
31:27
the key here is friction rates. You still want to make it fun like Miguel
31:32
said, right? And so, uh, the ability to use that type of a AI to drive
31:38
refinement across the player journey, whether it be step up notifications
31:43
versus hard blocks or warnings versus, you know, again, banning, that is
31:49
really, I think, the power of where AI can really drive profit.
31:53
>> Well, and Andrea, Miguel, actually, maybe Miguel, I’ll turn to you because
31:58
you’ve said slightly less than Andrea, I think. Um where do you feel you are with
32:03
with using AI in in this journey?
32:09
>> Um all is a bit hard to assess that uh due
32:13
to lack of data to properly compare it but I think we’re here at least we’re at
32:18
a good stage of using it. Uh the same way in the this panel, keeping
32:25
a healthy balance between the player protection and the the the margins. I
32:31
like to keep a healthy balance between the usage of AI while uh keeping a good
32:37
uh human control over things
32:40
>> uh at least until we’re very sure that the tools can properly function uh
32:46
without much intervention or much review from a human perspective. But I I tend
32:53
to be a bit pushy for the human side as I was me
32:58
mentioning a few minutes ago regarding the the human interaction because the
33:03
machines can indeed check the the logs of communications with the player but
33:07
there’s always nuances uh and picking on the localization uh there’s multip
33:16
multiple ways of saying the same thing and Andrea shares the the language with
33:20
me which is Portuguese, we tend to have a ton of ways of saying exactly the same
33:25
thing or the opposite. The same phrase different on different contexts can mean
33:31
very very different things for good and for bad. Um
33:37
this on a risk perspective for the players and this as well on a
33:41
perspective for fraud. Uh I was a bit ago I was thinking about the a random
33:46
example like in English you have the word thank the the two words actually
33:51
thank you in our case both can use it but me
33:56
andrea we would in Portuguese we would say differently actually Andrea would
34:00
say oblriata I would say and these are kuances that usually require a human to
34:08
look at this is a woman a woman woman’s account but It’s saying, “Oh my god,
34:14
doesn’t match. Something’s wrong here.” Uh, and I I tend to be a bit pushy for
34:20
the human side on this. uh the the the user of AI and machine learning are
34:27
incredible things for scale because you get to a certain point that you would
34:32
need a a whole warehouse full of risk and fod agents and another whole
34:38
warehouse full of responsible gaming agents to to handle a whole casino. So
34:43
again, AI is crucial for to scaling up, but you always need that bit of human
34:50
check on things to to properly assess those nuances, those details that can
34:56
be, let’s be conservative, can be very very
35:02
hard to tackle on the on on with AI because they’re so nuanced, so niche
35:10
>> can be very tricky. Andrea, would you Oh, go on, Alex.
35:15
>> I want to double click on that because I agree entirely. And there was a saying
35:18
that came out uh 10 12 years ago. I forgot who said it, but I remember
35:22
reading it or hearing it somewhere. And what they said, and this is of course
35:26
prior to AI and ML really taking off across the industry. What they said was,
35:31
“You can only automate what you can reliably predict.” Well, in all lines of
35:37
our work, what can we predict? You know what I’m saying? Uh and so the
35:42
limitations of AI are real and the goal of accuracy across AML, across uh
35:51
compliance items, across regulations and fraud. The key point is to be accurate
35:55
so that we can be revenue generating and we can sustain growth and all of these
36:00
different things. Accuracy is key. And without that human element, if we were
36:03
to solely rely on AI at different items for things that cannot be predicted,
36:08
yes, we’re in for a world of trouble. some ago. I just really wanted to double
36:11
click on that.
36:12
>> Would you, Andrea, would you say that in different parts of your role it’s more
36:17
useful? Like I don’t know in fraud protection or in player protection or in
36:23
AML. Did would you say that your AI has been more important to you in in one bit
36:30
rather than another?
36:33
>> Yes. Uh I I subscribe what uh everyone on this panel is saying about AI. Um
36:39
what we have observed is that uh AI is transforming the way that we managing
36:43
risk right now and and will be affecting more in the future. But from my
36:48
perspective we have to use it responsibly. It help us to connect data
36:53
scale.
36:59
>> Oh. Oh no.
37:01
>> Oh no. And you’ve frozen.
37:05
>> The weather is really bad in Lisbon. um the power of um
37:10
>> sorry Andrea we lo we we lost you for a moment there perhaps you could rewind
37:15
>> rewind 30 seconds and tell us again what you’re saying yeah yeah yeah
37:19
so I was I was mentioning I was subscribing the what everyone on the
37:23
panel was saying about EI uh that EI is transforming um what we how we manage
37:29
risk and a scale ways but my perspective is that we use need to be used
37:39
that we cannot um combine um a behavioral data in one
37:47
prop player profile and also it will help Can you hear me? Can you hear me
37:52
well?
37:52
>> Yeah, we can now. Yeah.
37:53
>> Okay. Um so the power of AI also lies great probably is the the massive rain
37:59
in Lisbon today. Um so the power also of VI is also lies in early detection not
38:06
in automated decision making. So from in perspective we focus on explanable AI
38:12
models that we can audit and challenge but we need to include transparency and
38:18
ethics. This is important as an accuracy. For example, I remember using
38:24
there are some cases where we face multiple um low-risk profiles uh with
38:30
the same payment instrumentment uh not criminal per se of course but if
38:37
we link to a shared gambling account um in one household or even other factors
38:42
can be a trigger. So behaviorally we are moratory issue.
38:52
So
38:58
>> my my opinion here is that yeah I
39:05
>> Okay, Andrea, we’ve lost you again.
39:07
>> We lost me again. Can you hear me?
39:10
>> Yeah. Oh, we can hear you now. I think we’re getting the gist. I might
39:14
I might move on if more. But no, no, you’re clear now. Do
39:20
you want to just finish what you’re saying?
39:22
>> Yes, I I finished. I have finished. Okay,
39:25
>> I finished. Let me just refresh my page. See if everything.
39:28
>> Okay. Um Stephanie, if your dog’s not going to introduce
39:35
He seems happy. What I would love to um kind of pick Miguel’s brain on and you
39:40
know if Andrea is able to come back is that what we talked about today are very
39:44
big theme things big organizational changes and TAN is a huge organization
39:50
and so so is yours and if there are operators out there let’s say that don’t
39:56
have a really big you know um let’s say want to cultural shift or uh ability to
40:04
kind of drive that that shift at that level. Is there any advice that that you
40:09
would have in in terms of how to create and um remove these kind of silos
40:15
so that you know um things like AI can flourish with a more human um focus?
40:24
>> You go.
40:25
>> Yeah. Yeah. Probably I’ll try to to help a bit on that. Uh not entirely sure if I
40:32
correctly understood the question, but I think I got it. Um
40:36
I think a good way would be to start using it as um very efficient um trainee
40:45
helping you scaling decisions or even having it as your
40:53
I would call it like that but your good side versus your bad side like
40:57
challenging your own assumptions on data challenging your own
41:02
interpretations of the data to see where it leads you. Could lead you to some
41:06
good decisions. Could lead you to a a confirmation of what you were already
41:11
thinking. Either way, you’ll be getting more information, different opinions,
41:18
which could help you start scaling things.
41:22
Could be perspective if I didn’t understand your question.
41:26
>> No. um you’ve had a different answer, but what what a a great response um
41:32
about how to kind of get started with with things like AO or the culture
41:36
trained. You know, Andrea Entain is a huge company um with with lots of
41:41
resources and the question to Miguel was kind of, you know, if folks don’t have
41:46
that kind of resources and a leader like you, h how does one kind of kind of go
41:51
about both this cultural shift? any practical advice or you know the
41:55
adoption of AI.
41:58
>> Um this is a really I I’m not there was some cuts in my in my in my in my uh
42:05
access here but I think everything is okay right now. This is a really good
42:09
point Stephanie indeed in is a big huge company with a lot of tools and and
42:15
systems. Um, EI is of course uh is it’s present on our days but is been used on
42:22
a safety and sustainable way. Um and also just for for example also it’s
42:28
important to have in mind that anyone anyone that joins in pass training
42:33
exercise from AML responsible gaming AI tools all of this information make sure
42:38
that we are um fully engaged everyone is fully engaged and everyone talks about
42:43
it u openly and not just about the risk but proactively um it’s not a just a
42:49
tickbox information um but is sure that everyone is fully aware ware of the
42:54
risks of the issues that it’s happening in the market and evolves as well.
42:59
>> Yeah. Um there’s something that we are we are coming towards the end of our
43:03
session but there’s something Stephanie that you and I talked about the other
43:06
day that I I wanted to to bring up in the conversation. Um you talked about
43:12
having checks. I think the traditional way of dealing
43:16
with AML is at KYC and and registration. Um but you were talking about um
43:24
monitoring players throughout the customer journey. Is that right? Perhaps
43:29
a bit on that.
43:30
>> See what a great point to wrap everything up. I didn’t even recall
43:34
having that that conversation. But yeah, that’s exactly kind of the point. And um
43:38
the crux I think is a is a happy bow to kind of the themes that we that that the
43:43
panelists have been able to share is that in my experience uh a lot of the
43:48
regulatory exposure or let’s say the um initiatives have really been focused on
43:54
implementing kind of standalone really robust front door checks whether it be
43:58
KYC and AML. And um what what the thought is though is that there are ways
44:05
that folks are kind of getting through that. It requires a lot of upfront
44:09
integration and um cultural changes just to get that um part of the approach
44:16
fortified. um Andrea talking about a layered approach is that a lot of the
44:21
behaviors that we kind of mentioned in some of the um examples that were given
44:26
were really kind of spikes in behaviors that happen post login. And so when we
44:32
see this mergence of KYC, AML, responsible gambling, promo abuse, all
44:36
the different use cases, collusion, you know, um really more complex use cases
44:42
that that touch all the different departments. Um it is increasingly
44:47
important um as we go through these real life examples to understand that the
44:53
initial investments up front while critical are simply kind of not enough.
44:57
the concept of a layered approach um a fortified approach at each part of the
45:02
player journey I think is is is very very key and I think the ability to do
45:07
so and driving profit is important because like Miguel said about being fun
45:12
and sustainable growth without that you’re kind of relying on a butcher
45:17
knife right to have hard blocks and then you know the revenue team will come to
45:22
you and say you’re blocking too much I can’t have this and so um having incight
45:27
um around those player behaviors, those dashboards, that cultural um
45:31
conversations that happen between departments allow things like AI to
45:36
automate um very more refined friction further downstream. And so you’re going
45:41
from a butcher knife to more of a a razor knife. Um and um in doing so, you
45:48
can drive more sustainable growth and allow players to still have fun.
45:52
>> Oh, fantastic. Um well before I bring things to a close, does anyone else have
45:58
any other observations they they’d like to share at this stage? This is your
46:03
last chance, guys. Um so that’s just left then for me to say
46:09
thank you to all our panelists. Thank you to Stephanie and Alex at Sift and
46:15
thank you Sift for your uh support with the digital day today. Um, thank you to
46:22
our panelists from Lisbon, Miguel and Andrea. Um, we managed to get through
46:27
the technical problems just about. It’s the first panel I’ve ever had a dog
46:31
attending that I that that I’ve known about. Um, and just thank you all for to
46:40
all the presenters and all the panelists in the in the other sessions. It’s it’s
46:44
been a bit it’s been a really good day. It’s been full of fascinating insight
46:48
and and I hope everyone’s managed to to glean something from it that that they
46:53
take home. So, uh, with that, um, I’m going to say, um, good night and thank
47:00
you very much.
47:02
>> Thank you. Bye-bye.