Andreas Welsch 0:30

Today we'll talk about what happens when your AI agents start transacting on your behalf and making financial transactions. And who better to talk about it than someone who's been actively researching this space for actually quite a long time in financial services and the consumer payments space, Michael Nevski. Hey Michael, thank you so much for joining.

Michael Nevski 0:49

Thank you, Andreas. I excited to be here today. Thank you for having me.

Andreas Welsch 0:53

Wonderful. We for those of you in, in the in the audience, we recently got connected through an, an event where we're both participating around consumer experience, and I thought, hey, I definitely want to have Michael on the show and also bring his insights to you. So Michael, I'm so grateful to have you on the show and to get some of your time because I know you're very busy. But for folks in the audience who don't know you yet, maybe you can share a little bit about yourself, who you are and what you do.

Michael Nevski 1:18

Thank you. Yes, I have over 15 years of marketing research experience in financial services as well as syndicated consumer packaged goods data, point of sales data, and data monetization. And in my capacity, I have done as a person who was responsible for global thought leadership in terms of the consumer trends, have done lots of research in the payment space as well as as well as wealth management and banking space. And I'm also very much involved with the research community on both sides of the pond in Europe and United States. And like you mentioned very often I participate in global conference events and learning and contributing and understanding what consumer preferences or how consumer preferences are changing as we adapting to new technology or any other events in the business in the world influencing consumer tastes and preferences and how they make payments, how they travel, how they consume products. That's my background, always partnering with the marketing product strategy. Another few area where I try to contribute in terms of the data monetization and how businesses, brands can leverage their data, whether it's transactional data or consumer data, to create a business intelligence which influence the relationship with consumers, influence the company's bottom line, and identifying those right spaces for the future.

Andreas Welsch 2:49

That's awesome. I'm so excited again that we're able to have this conversation today, and we obviously talked a little bit before going live here, so I know it's gonna be a great conversation. But you mentioned we want to find out more about consumers, consumer behavior. And I'm wondering, with everything that I'm seeing and reading everything that the AI labs like OpenAI put out there, you can shop through ChatGPT. I know a couple months ago they had a pilot with Walmart and made some changes to how how they're doing the transactions and what they're displaying within ChatGPT. But it seems like discovery, what is available that we've already talked about on, on the show in previous episodes, and now checkout are just two sides or two parts, I should rather say of that process. Finding out what's available and then saying, "Yes I want to buy this." Where do you see this going? Is this primarily a, a B2C topic? Is this a B2B topic? And where, yeah, where does it stand in, in, in the financial services space right now?

Michael Nevski 3:50

Yeah, absolutely. I think this is a great opportunity. If we look at the numbers nowadays it's almost ubiquitous in terms of the LLM or generative AI, how we call it. And everybody knows those, Anthropic, OpenAI, those guys which actually brought those first LLM models at the consumer level where people could just experience them, sign up for paid plans and stuff like that. But the next iteration, what you're referring to, is a agentic e-commerce, so agentic AI per se. And when I said the biggest awareness that many companies in the payment space been using AI, overall AI, right? Artificial intelligence for fraud, for risk, for a long time. And my studies, most recent studies from this fall from the spring and summer 99% of consumers in the US, for example, saying, "I'm very well aware of that." But when it comes to generative AI after almost three years it's about 80%, every four, four out of five, not every, four out of five consumers saying, "I am aware of generative AI. I understand and I know what it is." But next iteration, like you said, is coming that kind of a convenience, that kind of a reliance on the technology which is the agentic. And what it means if we kinda define that and say what is that. Agentic to me refers to AI system designed to act autonomously, making decisions and taking actions to achieve specific goals rather than merely generating outputs based on the input, right? The systems perceive their environment, reason about actions, and then execute those actions often in a multi-step process. This contrast with generative AI, which we all know, which primarily provides outputs that require further human action. And- That's what's called agentic e-commerce. And especially in the payments and the banking and the wealth management, many brands see that as a next opportunity a new, a next white space. Why? Because it will provide efficiencies and time savings for consumers. For example, if I am doing Staples and I need to buy toilet paper every month or something, I don't wanna spend time even going anywhere. I can just dispatch my bot, my agent create the order, and tell them the quality, the brand, the price, the retailer, the channel, whatever I want, and ask the bot to go and execute. And then it can be delivered to my home, right? Pretty much at my doorsteps. And that's the convenience. So I can spend more time with my family, I can be more efficient and buy the most up to... for the most optimal price and it creates a lot of efficiencies. And that's what agentic e-commerce. But to your point, to your question, when that transaction would happen right now, it's just in a sense potential. Two major networks, of course, already a year ago announced, made the announcement that they enabled the networks to payment networks to adapt to agentic e-commerce. But in reality yes, we've seen some adaptation by Walmart and OpenAI to create some of the elements of that, but it doesn't exist yet. And so what we're dealing with, we're dealing with a hypothetical use cases right now. And when that transactions happens, lots of questions arises right away. The privacy, the security, the chargebacks, the disputes, all of those needs to be resolved and solved for before the true network's gonna be enabled. Not only just the payment tokenization and security of that transaction, but all of the other aspects what I just mentioned, and then a lot more which involved in the whole equation. And overall, yes, agentic e-commerce is coming. In my opinion, it is inevitable. When you ask consumers, would you use it once you know you can rely on it and you build the trust, almost half of them saying yes because I see the benefit. But proof is in eating, right? So prove it to me as a brand, as a technology. Prove it to me that I can rely on that system. And that's the kind of, for consumers, jury's still out up there

Andreas Welsch 8:17

So let me jump there because that would have been the question too. How realistic is this? Is this conceptual at the moment? Are there tests? Are there evaluations? Yeah. Is this live in production? But I think you already answered that to the degree saying the, the concepts are there, but now it's a matter of not only prototyping, but implementing this and implementing this along the entire chain, right?

Michael Nevski 8:39

Yeah, you're absolutely right. In a sense, concept is there. There is a agent verification protocol already in place. So that means if I dispatch the bot, a system or network can identify authentically that it was my agent. There are some credentials which can be stored and tokenized and verified. You There, there are also techniques already in terms of the agent answer optimization. That means information about the product needs to be clearly stored, accessible, definable, so the bot can really go into your merchant website, identify that easily, and compare apples to other offers, other providers. So all of that information, our system, in a sense already is evolving and may the prototype level in a sense is available. It can be enabled. But there is another question the other side of this equation consumers and sellers, right? Why? Because sellers needs to really adapt, like Walmart is doing what we mentioned earlier, right? So with that AI system, but not only AI system, but actually, yes, making those purchases within the, LLM app, which again will dispatch that bot to create that transaction. And so majority of merchants are still in a sense of that adoption curve of early adapters. Not early majority, not late majority, but really early adapters and innovators. They're still looking at it as kind of a new technology and wanna see more proof that technology will work for them. And the second part of this equation, consumers, they also need to adapt. They also need to start trusting this system in the sense of technology. And that's what most fascinating right now because the- any system can be built, technology can be provided, but would Like, in that expression, build it, they will come. Would they come once it's built? And that's the conundrum.

Andreas Welsch 10:38

So thinking back 25 years roughly. When eBay was one of the major platforms to buy products or especially secondhand products. I remember there, there were a number of sniping tools that would alert you if the product became available at your threshold price or below that, and then they would purchase on your behalf. So we've seen this a, a while ago. Now we have services like, what's it called? Camelcamelcamel that alert you if if there's a price drop on Amazon. It's not sniping. You still need to perform the transaction yourself. Amazon last last fall filed a lawsuit against Perplexity who are coming out with their agentic browser saying, "No, you're not allowed to have that agentic browser browse our catalog because, hey, we're still making money from ads, and so we don't want to be disin-intermediated. We still need to show ads to people." And then on the other hand, you have services from Amazon like Save and Subscribe or Subscribe and Save. You say, "Every month I need dish soap. Every month I need toilet paper," and it just shows up on my doorstep on, on, on a certain day. So in a sense, I feel as consumers, we are already conditioned to delegating some of that agency to an, to a system It seems that it's always been more role-based. If this happens, if the price drops below this amount, then do something in very narrowly defined actions. But now it's, "Hey find me the best the best cordless drill for this task, or give me three options, and then buy the one that shows up tomorrow morning on my doorstep because I need it tomorrow." So i- there's more variability that comes into it, and I think they're probably placed to... That, that fact that you mentioned about the consumers getting more comfortable with that side of it because the other ones I feel we've had for quite some time.

Michael Nevski 12:26

Yeah, you're absolutely right because there is a kind of a much bigger element of autonomy on behalf of bots. It's like executing selling or buying stock, right? So when price drops below a certain threshold, you say, "Buy me stock or sell my stock." Same idea with the eBay or prices, so they often will alert you either, and would buy it for you if you put in that put order, or sell it or doing something. Here, you're just pretty much dispatching the bot, and especially for something because you can set it and forget it, and then maybe you don't need that anymore a week later or two weeks later, right? But here, you're dispatching a bot for, call it instant gratification, but you needed those staples now, right? Within a day or two, so you're not gonna forget that because it's something you consume. And I'll mention some other use cases, but right now let's stick with- traditional shopping, right? And then you pretty much let it go because you don't know exact price. You don't specify, or maybe you can, but same time you want to optimize it. And a quantity, and a quality, and a brand. So many aspects of our decision which we usually control is given to machine and associated technology. That's why right now if you ask that question which we did about how much would you feel comfortable to spend- via this technology, agentic e-commerce via agent, if that would exist right now, hypothetically. And guess what people saying? No more than play money in Vegas, 10 to $20. Yeah. Because it's almost like, "Okay, I'm ready to lose it, so it doesn't matter," right? And the reason why I just wanna build on that, because if you look at consumers' kinda thinking right now, mindset, what are they thinking? What they're thinking that on the positive side it's about personalization and context to where decision-making, efficiencies, automations, innovation, definitely. It's very positive. Positive social impact because we can not only shop, we can manage smart homes, and that means, for example, not only just turning on the lights at night or opening garage door, but also ordering the supply, like water filters or some supply for your refrigerator or stocking for your pet in terms of the pet food or something else. Or turning on the lawnmower, right? So those are autonomous electric ones. So pretty much it can be your butler almost, right? Taking care of your house, right? But same time when they're thinking about the concerns, it's about lack of human oversight and control. Because still we as humans, we're gonna custom to control those transactions. It's also about trust and reliability. Who is responsible for what? What if transaction goes wrong? So what about the privacy and security? Is my data g- data gonna be safe secure, stored? So but also even a technological skepticism. Can it really do this what you're telling me to do, right? So and where it leads us, this is more of, logical, rational thinking. But also I try to uncover that one layer below by asking s- certain questions, right? To understand what about emotional aspect of that, what humans are thinking about it. And definitely the biggest word in my cloud was confidence. Wow, we were really starting flying to the moon, and now we're like com- coming up with this technology, and optimism, curiosity. I'm gonna be excited. I'm gonna be empowered to do something like almost like pretty much relying on technology like in the sci-fi movies. But same time, people still don't understand the thinking. Even early adopters. Can it really do that for me? How does... What's the black box behind it, right? So how the system, in terms of the purchase process, selection, right? Touch and feel. How is it gonna be? And when we don't know something or don't understand, what's interesting, the biggest factor is fear, because we falling back on our primary kind of feelings, like in the medieval ages when we don't understand something, like the gods, and stuff like that. So we're fearing. We're having skepticism, even distrust, concern. And those factors needs to be overcome before this agentic commerce can be adopted by consumers, because they need to make sure, they need to build that trust. And another interesting data point for you 77% of consumers saying that they anticipate agentic commerce and generative AI will improve their lives. But the very same pretty much consumers, 83%, because they overlap so much, they're having a concern about AI in terms of the oversight, in terms of the reliability. And so what I call it, right? So high hope, high fear, kinda. This love-hate relationship, if you call it, right? Because on one hand, hypothetically, I'm so excited. The same time, I don't understand, and then think about it, I'm so scared, right? So and that's in limbo. That's where consumer mindset right, right now is

Andreas Welsch 17:38

Boy, that's a, a big paradox. It feels like the, the excitement and at the same time the anxiety about what is happening, and I cannot really grasp it, but it seems great, but it's also really scary. Now one, one question that I wanted to ask you, or actually two in, in that same direction is, sometimes you might get a call from your bank or a text message, "Seems that there is a fraudulent activity. Can you confirm these transactions? Were they performed by you, yes or no? If yes, all is good. If not, maybe we'll lock your card or we'll send you a new one." Or maybe you make an impulse purchase at 3:00 in the morning when you lie awake and you s- doom scroll Amazon, and next morning you regret it. What does that look like eventually when agents can transact on your behalf? Meaning who or who is responsible? Is there a way to wind this back and say sorry, I didn't mean to buy this. Can I get a return?" And more importantly what happens when the agent goes off the rails and buys something that you didn't want, didn't need, that is extremely expensive, that you might not even have the funds to pay for Who authorizes the payment, and who's responsible and accountable at the end for making this purchase? How do you troubleshoot that?

Michael Nevski 19:10

Yeah. And in this equation, the short answer to authorizes is still should be your issuer, the bank which issued the card. They will review that transaction in this moment instance and they look at all the credentials. But the, that's the short answer in the sense of who authorizes. But really you have You ask a very good question. What consumers think, and I will go I'll set the stage. I ask this question about bot technology, not necessarily that type of AI, generative AI overall and stuff, but bot technology specifically. And what's interesting, although you would need to take those numbers with a grain of salt because they're close. But when I ask that question, who would you trust the technology provider once you know the technology works and stuff like that? And in general, 58% of consumers saying, "Yes, once I feel confident, once I understand and know how it works and who's responsible for what, if a technology provider gives me this opportunity to sign up for this bot to shopping, this bot to deal with my health or this bot to manage my house, for example, I'm fine." But when I pivoted that and asked, "What about if retailers or merchants start providing that service?" Instead of 58, that percent dropped to 52. Yes, six percentage points, about 10%, 11%, it's not significant, but at the same time it tells us that consumer mindset that I want to deal with retailer as a retailer, as a seller. Stay in your lane. Sell the products and services to me, not the technology. And why I'm mentioning that, because again, back to your question almost 40%, 39%, two out of five consumers feel that if transaction goes wrong The agentic AI service provider is responsible for that transaction. Yes. But what's interesting, 6% only said that it would be the payment networks, right? So we're not gonna call those names, but we all know who they are. And 17% saying actually it's AI developer not the bot provider, but the AI developer like Anthropic OpenAI, those guys, because they're still looking for those specialized providers. And even almost one out of six, 14% saying it's also on me because I might be wrong as a user, which was a big surprise to me that one out of six saying I could create the wrong order, purchase order or something, so I maybe made a mistake, and stuff like that. But also consumers want the system, ecosystem having a rails, right? So network, having issuers, merchants and merchant banks all the parties and bot providers to have that role and figure out how process is gonna be working for consumers, right? So- and that's what we need to keep in mind. Yes.

Andreas Welsch 22:00

Yeah. But so I think this experience, and especially the customer experience part gets so much more important because as a customer, depending on who and what you interface with, if you have a personal agent that acts on your behalf, that sources information and gives you the, the top three or five products that you should buy that meet your criteria and that does the checkout for you, or if you go to a vendor's website and use their agent on their property I'm assuming there's a, a perception and maybe even a negative perception if one of those links in that chain breaks, or it takes longer than expected, or it goes off the rails. Doesn't have to be just the payment network. It could be a catalog item in your vendor's e-commerce system that isn't properly maintained, so you don't get the best recommendation, or if parts of the checkout process take longer. So it seems that there's a lot more to think about. Not necessarily new things, because I think even, a, a delay in your checkout process today if the little hourglass or wheel keep spinning, gets a little annoying at times. But now there are different actors and not everything is human in the process. But for, for our final question I'm really curious, what do you recommend business leaders, IT leaders, marketing leaders start doing right now to prepare for this reality? And finance leaders should probably look at this too, huh?

Michael Nevski 23:18

Yeah. Andreas, that's a very good question, but before I answer that, I have a one comment on the previous topic. Yeah. Actually, you're absolutely right. It's not only experience, it's also consumers understanding value of their personal data. To give you an example when I asked the question, would you share the, your data, your private data, or your purchase data, your behavioral data with bots, right? To make them, to enable them to be even smarter. The numbers were overwhelming just the, for example, location data. Only 7% said, "I would share location data just in general within my app," or internet browsing, 6%. But however, when I pivoted that and I said, "What about you're in the middle of the city and you need to find a restaurant for business lunch within an hour, within one mile?" The percentage of location data jumped from 7 to 45 When internet browsing from eight to 43. Because if it's a specific use case, consumer say yes, because I need to solve this, right? So it's like a, I need to make a hole in the wall, not a drill, right? Like the classical case from Home Depot I remember from business school. So consumer saying my data is valuable. Yeah. And if I need to accomplish this particular task, with that particular task, for this time frame, I'll give you my personal data to find that restaurant. But not in general. I don't want you to be Big Brother. And so that's why it's very important to understand the checkout process, the autonomy of those bots, right? So consumers who are willing can open and understand the value, but they wanna get a trade-off. They wanna get something for it. And moving on to your last question- Yeah about what leaders, IT leaders, and business leaders needs to do. So a few things, right? So definitely we need to understand who are early adapters. And we didn't talk about it, but of course the younger consumers. So we need to start from understanding your customer, understanding your target audience. That's the Gen Z and millennials. But you need to also enable, whether it's merchants Or issuers, they need to start thinking how they're gonna enable and participate in this ecosystem, right? So the payment ecosystem to enable agentic e-commerce. That means for merchants to start adapting to agent optimization an agent answer optimization protocol or with agent verification protocol, or start listing pretty much, in my opinion SEO. So changing optimization is gonna be saving back significantly because sites will be optimized for bots in the future, for them to clearly define and see their products and services and qualities and ingredients all clearly listed and stuff like that. But also issuers as well to enable the payment network, the cards and stuff like that, tokenization. And also we need to provide tools for consumers for easy opt-in and opt-out controls. Opt-in controls and opt-out if they don't want to. We need to also think of from the technology perspective, what legacy systems I have. Do I need to modify that, upgrade adapt to new ones to enable that in the future? We need to also collaborate with governments and regulators, not only in the United States, of course, but worldwide, right? Because this is coming, and especially again, younger generations who are mobile native, digital natives, they will rely more and more on those systems because me and you probably grew up in the era when we didn't have really mobile phones in a sense, right? So when we st- became a little bit older, probably it was a kind of a ubiquitous. But, new generations, for them, they never knew what a regular dial-in kind of phone is, right? And stuff like that. And the last one, probably we need to continue listening to consumers because tastes and preferences and trends evolving among consumers and generations and segmentations. So that's why we need to really understand the market and continue be close to that.

Andreas Welsch 27:23

There's a long list. So pretty much everyone needs to get involved no matter what role you're in. If you're in IT, if you're in finance, in marketing, in supply chain you need to start thinking about this even if it is something that you're not implementing tomorrow, but being aware of where where the puck is moving, right? To use a hockey an- analogy, so you can skate there and make sure that you're in the right spot at the right time once this matures. Michael, I, I've learned so much over the last 30 minutes. I can't believe we're al- almost out of time. Thank you so much for sharing your perspective. I know there's a lot more to cover. Maybe we can do that on a future episode or folks can reach out to you on LinkedIn. But I'm so grateful to have had you on the show, and thank you for all your insights.

Michael Nevski 28:03

Thank you so much. That was my pleasure, and I really enjoyed our conversation as well.