Hi everybody and welcome to a special episode
of The Saturday Fraud Strategist. Here live from Money 2020 Vegas.
With me I have Ben Herut VP of payments at Chargeflow. Welcome Ben.
Thank you so much. Thank you for having me.
It's always a pleasure to be there with all colleagues.
Yeah. Uh we have a special bearded episode uh today.
Yeah. Uh Ben, so I wanted to speak to you about agentic commerce.
Um so I you know I spent some time looking into this.
I wrote even a a paper about it uh I think
about like 6 or 12 months ago.
But you're one of the experts in the industry.
Uh so I wanted to grab you and ask you a couple of questions.
And you know like let's start with concerns.
Uh there are a lot of open questions about uh agentic
commerce at the moment, uh and a lot of unknowns.
And I'm just wondering you know when you're looking at the space
right now what are the biggest question marks that you are concerned
about or worried about that are not answered yet?
Well I think the biggest one is who's liable I think nobody actually
knows I think Visa and Mastercard and AMX are trying to get there, right?
But as always, it's a step back.
And at the end of the day you're throwing
into the equation a wild card, right? A joker.
And nobody actually knows how it's going to perform.
Obviously, there are testing and so on, There are specific partners.
But at the end of the day, someone is doing actions that were
made by humans in the past, and now we need to adapt.
And the whole ecosystem needs to understand
that now there's a new player in town.
And how do we how do we define who who's at fault, who's liable?
Because I think it's also like we're not officially, but sort
of, introducing a new actor into the payment chain, right?
Which is the actual AI provider, right?
Um, so are they liable at certain cases?
Why is it just the merchants or the issuers?
And usually when I talk about agendic commerce, I compare it to BNPL.
Because it had the same type of influence on on the industry.
Because at the end of the day, you've got the BNPL
companies themselves, but they were never liable.
Was it It was still in between the issuer or the merchant, right?
The liability shifted based on the contract that they were signed.
But here again, you've got the agentic commerce.
Are OpenAI going to be liable for something?
Is the issuer going to be liable?
Mmm, from my experience probably not.
I guess it depends on the type of uh information
that run through the system.
But again, the biggest issue is that nobody actually knows.
And and I saw that, like to a degree, it seemed like earlier in
the year that the schemes are apprehensive around uh sharing
with issuers um whether a transaction is agentic or not.
But it seems that they walked it back to a degree and now they are.
Uh but it's not really clear how and when and um
and what issuers would be able to do with it.
Like how do you see it at the moment?
What what can issuers when they look at agentic commerce and you know
that that's something that for them is very hard to deal with, right?
Uh you know they haven't been or they are not in
direct uh uh contact with the the shopper.
Um and even with 3D secure um when know 3D secure you
can relay the device, uh fingerprint, and so on.
Even that is now no longer in the game because you have agentics.
How do you like how would issuers uh play?
So again at the end of the day uh when we started talking about AI uh
payments right that they usually spoke about three different mandates.
You got the payment mandates, the card mandates, and uh you know
the basically the approval itself or the intention mandate.
Um as long as you are not sharing this info it's going to be very
difficult for anyone to do, to make a decision on top of it.
At the end of the day if I own agents right and I I operate them.
I allow them to buy something on my behalf.
It has to be not only clear what I'm allowing them to
do, but also it has to be registered on both sides.
So the issuer knows this is an agentic commerce.
I need to be slightly more careful, potentially.
And also once it's hit the merchants, right?
They need to be able to show that they are
interacting with a with a with an agent.
So they have to make sure that their storefronts are available for
AI to easily skim through and find whatever they need to find.
So okay, let let me let me uh pick up on this.
So as an issuer, you know, one of the main thing I
I mentioned it uh a second ago, one of the main tools
that you have to control fraud is 3D secure.
Um but 3D secure you know assumes that you have a human in the
loop meaning a human uh uh uh placing the transaction.
When an agent does that it might be that
the human is is not even in the loop.
Now now we have uh these these Muse and Dots and all of these
kind of like assistants AI assistants that won't necessarily
even alert you before they base a transaction.
So, a human might not be, you know, like
aware of their phone or whatever.
I always give the example, maybe they're
in at the dentist at the moment.
Maybe they're on a plane, maybe they're asleep.
How like how does 3D secure uh come into play here?
Like in my mind, it's it's you know, it's broken.
Uh how how do you how do you uh introduce authentication
flows in agentic uh commerce flows?
So that's a very tough question to be honest.
And I I'll take it even a step further because the whole chargeback
or dispute ecosystem was built on human behavior.
None of it was ready to to to have this new actor in it.
Um 3D secure is one one of the aspects.
But as you can see for example we've got the PSD2 that transferred
into the PSD3 and they made the relevant changes into it.
But at the end of the day they would find a way to make sure that the
the new data, or the new performance of how each one of the actions is
taken on the ecosystem, would be monitored and taken into consideration.
As far as I know, there's no there's no one answer currently.
But it's it's still believe and I think it's very important to
look into and hopefully we're going to have more info soon.
It's changing on a weekly basis almost.
Yeah, that's super interesting.
Let me flip it now and ask about first party fraud.
Because I I think with third party fraud, I mean
I understand the the the issues or like the how
Yeah, the complexity. And how the chain breaks.
But with first party fraud, it's you know it's it's a bit bizarre.
We we are very focused on AI hallucinations and whether they can, you
know I I you know once gave the example you know buy me an iPhone and
they buy a picture of an iPhone right as an as an as an hallucination.
But what happens like what are the different scenarios that
can happen with first party fraud and agentic commerce.
And how should merchants or issuers uh think
about it as as they try to prepare for it?
So I I don't want to have anyone too scared
but basically it opens a Pandora box.
Because until now we I mean the the the basic statistic is
70% of chargebacks are uh friendly fraud or invalid.
And these are usually in in two different sections.
One of them is is uh unknown real unknown uh uh dispute
reason, and the other one is intentional. Right?
So now when you add the layer of complexity of the someone else
interacting on my behalf, and I hope I made the correct prompt
to the agent to buy the correct stuff that I wanted.
Now you just a roomful uh basically denying
This was me. Is just that much bigger.
It's the it's the earth excuse for buy remorse. Right.
Yeah. I didn't mean that. It was, you know, it was a hallucination.
Absolutely. So, it just opens up a new world of denying, right?
Um, and this is a challenge that everybody is going
to encounter at a certain time in place.
I mean we know that nowadays agentic permits are just in
the you know starting, and it's not that wide common.
But it's going to hit everyone.
Which is interesting because we're seeing that,
you know let's take Amazon as an example. And Amazon is a trend setter.
They don't allow AI assistance To shop
Maybe they know something we don't know
But how do you guys look at the agent payments?
Agentic payments? Well I said completely the same.
Uh yeah, I mean there's the KYA aspect and yeah, there's like
the body detection aspect, but I think in my mind the the point
that you made about liability shift and where it lies.
And you know, liability in the end, it's you know who pays
is also who would invest in the solution, right?
So while we don't have that, and while by the way while
we don't put any motivation on the AI uh um labs or the
AI providers themselves, to invest in security.
In uh being able to you know like block uh ADO, block uh
stolen identities, block um stolen API credentials.
While we don't give them the motivation, We are introducing
a player that just wants to make money.
Uh and I don't I don't see it, I don't want to say I don't
I don't want to sound an apocalyptic as you said.
But I don't see it ending well with um with a setup right now.
And I agree with you that you know the schemes kind of
like are taking a step back to watch how things play.
And someone would pay the price for it,
which is probably the merchants.
Ben, it was a pleasure having you.
We should uh do something like proper on the pod itself.
Oh, of course. It was a pleasure. Thank you so much for inviting me.
Always a pleasure. And yeah, let's see what happens.
All right. Thank you very much. And folks, I see you next Saturday.